xref: /dflybsd-src/sys/net/if_ethersubr.c (revision 9b42cabedff59ea3b351e8eccb2da9bab5cefe7e)
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 1.75 2008/07/07 22:02:10 nant 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_mpls.h"
43 #include "opt_netgraph.h"
44 #include "opt_carp.h"
45 #include "opt_ethernet.h"
46 
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/globaldata.h>
50 #include <sys/kernel.h>
51 #include <sys/malloc.h>
52 #include <sys/mbuf.h>
53 #include <sys/msgport.h>
54 #include <sys/socket.h>
55 #include <sys/sockio.h>
56 #include <sys/sysctl.h>
57 #include <sys/thread.h>
58 #include <sys/thread2.h>
59 
60 #include <net/if.h>
61 #include <net/netisr.h>
62 #include <net/route.h>
63 #include <net/if_llc.h>
64 #include <net/if_dl.h>
65 #include <net/if_types.h>
66 #include <net/ifq_var.h>
67 #include <net/bpf.h>
68 #include <net/ethernet.h>
69 #include <net/vlan/if_vlan_ether.h>
70 #include <net/netmsg2.h>
71 
72 #if defined(INET) || defined(INET6)
73 #include <netinet/in.h>
74 #include <netinet/in_var.h>
75 #include <netinet/if_ether.h>
76 #include <net/ipfw/ip_fw.h>
77 #include <net/dummynet/ip_dummynet.h>
78 #endif
79 #ifdef INET6
80 #include <netinet6/nd6.h>
81 #endif
82 
83 #ifdef CARP
84 #include <netinet/ip_carp.h>
85 #endif
86 
87 #ifdef IPX
88 #include <netproto/ipx/ipx.h>
89 #include <netproto/ipx/ipx_if.h>
90 int (*ef_inputp)(struct ifnet*, const struct ether_header *eh, struct mbuf *m);
91 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst,
92 		  short *tp, int *hlen);
93 #endif
94 
95 #ifdef NS
96 #include <netns/ns.h>
97 #include <netns/ns_if.h>
98 ushort ns_nettype;
99 int ether_outputdebug = 0;
100 int ether_inputdebug = 0;
101 #endif
102 
103 #ifdef NETATALK
104 #include <netproto/atalk/at.h>
105 #include <netproto/atalk/at_var.h>
106 #include <netproto/atalk/at_extern.h>
107 
108 #define	llc_snap_org_code	llc_un.type_snap.org_code
109 #define	llc_snap_ether_type	llc_un.type_snap.ether_type
110 
111 extern u_char	at_org_code[3];
112 extern u_char	aarp_org_code[3];
113 #endif /* NETATALK */
114 
115 #ifdef MPLS
116 #include <netproto/mpls/mpls.h>
117 #endif
118 
119 /* netgraph node hooks for ng_ether(4) */
120 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
121 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp,
122 		struct mbuf *m, const struct ether_header *eh);
123 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
124 void	(*ng_ether_attach_p)(struct ifnet *ifp);
125 void	(*ng_ether_detach_p)(struct ifnet *ifp);
126 
127 int	(*vlan_input_p)(struct mbuf *, struct mbuf_chain *);
128 void	(*vlan_input2_p)(struct mbuf *);
129 
130 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
131 			struct rtentry *);
132 static void ether_restore_header(struct mbuf **, const struct ether_header *,
133 				 const struct ether_header *);
134 static void ether_demux_chain(struct ifnet *, struct mbuf *,
135 			      struct mbuf_chain *);
136 
137 /*
138  * if_bridge support
139  */
140 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
141 int (*bridge_output_p)(struct ifnet *, struct mbuf *);
142 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
143 
144 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
145 			      struct sockaddr *);
146 
147 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
148 	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
149 };
150 
151 #define gotoerr(e) do { error = (e); goto bad; } while (0)
152 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
153 
154 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
155 				struct ip_fw **rule,
156 				const struct ether_header *eh);
157 
158 static int ether_ipfw;
159 static u_int ether_restore_hdr;
160 static u_int ether_prepend_hdr;
161 
162 SYSCTL_DECL(_net_link);
163 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
164 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
165 	   &ether_ipfw, 0, "Pass ether pkts through firewall");
166 SYSCTL_UINT(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW,
167 	    &ether_restore_hdr, 0, "# of ether header restoration");
168 SYSCTL_UINT(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW,
169 	    &ether_prepend_hdr, 0,
170 	    "# of ether header restoration which prepends mbuf");
171 
172 /*
173  * Ethernet output routine.
174  * Encapsulate a packet of type family for the local net.
175  * Use trailer local net encapsulation if enough data in first
176  * packet leaves a multiple of 512 bytes of data in remainder.
177  * Assumes that ifp is actually pointer to arpcom structure.
178  */
179 static int
180 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
181 	     struct rtentry *rt)
182 {
183 	struct ether_header *eh, *deh;
184 	u_char *edst;
185 	int loop_copy = 0;
186 	int hlen = ETHER_HDR_LEN;	/* link layer header length */
187 	struct arpcom *ac = IFP2AC(ifp);
188 	int error;
189 
190 	ASSERT_NOT_SERIALIZED(ifp->if_serializer);
191 
192 	if (ifp->if_flags & IFF_MONITOR)
193 		gotoerr(ENETDOWN);
194 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
195 		gotoerr(ENETDOWN);
196 
197 	M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT);
198 	if (m == NULL)
199 		return (ENOBUFS);
200 	eh = mtod(m, struct ether_header *);
201 	edst = eh->ether_dhost;
202 
203 	/*
204 	 * Fill in the destination ethernet address and frame type.
205 	 */
206 	switch (dst->sa_family) {
207 #ifdef INET
208 	case AF_INET:
209 		if (!arpresolve(ifp, rt, m, dst, edst))
210 			return (0);	/* if not yet resolved */
211 		eh->ether_type = htons(ETHERTYPE_IP);
212 		break;
213 #endif
214 #ifdef INET6
215 	case AF_INET6:
216 		if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst))
217 			return (0);		/* Something bad happenned. */
218 		eh->ether_type = htons(ETHERTYPE_IPV6);
219 		break;
220 #endif
221 #ifdef IPX
222 	case AF_IPX:
223 		if (ef_outputp != NULL) {
224 			error = ef_outputp(ifp, &m, dst, &eh->ether_type,
225 					   &hlen);
226 			if (error)
227 				goto bad;
228 		} else {
229 			eh->ether_type = htons(ETHERTYPE_IPX);
230 			bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
231 			      edst, ETHER_ADDR_LEN);
232 		}
233 		break;
234 #endif
235 #ifdef NETATALK
236 	case AF_APPLETALK: {
237 		struct at_ifaddr *aa;
238 
239 		if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL) {
240 			error = 0;	/* XXX */
241 			goto bad;
242 		}
243 		/*
244 		 * In the phase 2 case, need to prepend an mbuf for
245 		 * the llc header.  Since we must preserve the value
246 		 * of m, which is passed to us by value, we m_copy()
247 		 * the first mbuf, and use it for our llc header.
248 		 */
249 		if (aa->aa_flags & AFA_PHASE2) {
250 			struct llc llc;
251 
252 			M_PREPEND(m, sizeof(struct llc), MB_DONTWAIT);
253 			eh = mtod(m, struct ether_header *);
254 			edst = eh->ether_dhost;
255 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
256 			llc.llc_control = LLC_UI;
257 			bcopy(at_org_code, llc.llc_snap_org_code,
258 			      sizeof at_org_code);
259 			llc.llc_snap_ether_type = htons(ETHERTYPE_AT);
260 			bcopy(&llc,
261 			      mtod(m, caddr_t) + sizeof(struct ether_header),
262 			      sizeof(struct llc));
263 			eh->ether_type = htons(m->m_pkthdr.len);
264 			hlen = sizeof(struct llc) + ETHER_HDR_LEN;
265 		} else {
266 			eh->ether_type = htons(ETHERTYPE_AT);
267 		}
268 		if (!aarpresolve(ac, m, (struct sockaddr_at *)dst, edst))
269 			return (0);
270 		break;
271 	  }
272 #endif
273 #ifdef NS
274 	case AF_NS:
275 		switch(ns_nettype) {
276 		default:
277 		case 0x8137:	/* Novell Ethernet_II Ethernet TYPE II */
278 			eh->ether_type = 0x8137;
279 			break;
280 		case 0x0:	/* Novell 802.3 */
281 			eh->ether_type = htons(m->m_pkthdr.len);
282 			break;
283 		case 0xe0e0:	/* Novell 802.2 and Token-Ring */
284 			M_PREPEND(m, 3, MB_DONTWAIT);
285 			eh = mtod(m, struct ether_header *);
286 			edst = eh->ether_dhost;
287 			eh->ether_type = htons(m->m_pkthdr.len);
288 			cp = mtod(m, u_char *) + sizeof(struct ether_header);
289 			*cp++ = 0xE0;
290 			*cp++ = 0xE0;
291 			*cp++ = 0x03;
292 			break;
293 		}
294 		bcopy(&(((struct sockaddr_ns *)dst)->sns_addr.x_host), edst,
295 		      ETHER_ADDR_LEN);
296 		/*
297 		 * XXX if ns_thishost is the same as the node's ethernet
298 		 * address then just the default code will catch this anyhow.
299 		 * So I'm not sure if this next clause should be here at all?
300 		 * [JRE]
301 		 */
302 		if (bcmp(edst, &ns_thishost, ETHER_ADDR_LEN) == 0) {
303 			m->m_pkthdr.rcvif = ifp;
304 			netisr_dispatch(NETISR_NS, m);
305 			return (error);
306 		}
307 		if (bcmp(edst, &ns_broadhost, ETHER_ADDR_LEN) == 0)
308 			m->m_flags |= M_BCAST;
309 		break;
310 #endif
311 #ifdef MPLS
312 	case AF_MPLS:
313 	{
314 		struct sockaddr *sa_gw;
315 
316 		if (rt)
317 			sa_gw = (struct sockaddr *)rt->rt_gateway;
318 		else {
319 			/* We realy need a gateway. */
320 			m_freem(m);
321 			return (0);
322 		}
323 
324 		switch (sa_gw->sa_family) {
325 			case AF_INET:
326 				if (!arpresolve(ifp, rt, m, sa_gw, edst))
327 					return (0);
328 				break;
329 			default:
330 				kprintf("ether_output: address family not supported to forward mpls packets: %d.\n", sa_gw->sa_family);
331 				m_freem(m);
332 				return (0);
333 		}
334 		eh->ether_type = htons(ETHERTYPE_MPLS); /* XXX how about multicast? */
335 		break;
336 	}
337 #endif
338 	case pseudo_AF_HDRCMPLT:
339 	case AF_UNSPEC:
340 		loop_copy = -1; /* if this is for us, don't do it */
341 		deh = (struct ether_header *)dst->sa_data;
342 		memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
343 		eh->ether_type = deh->ether_type;
344 		break;
345 
346 	default:
347 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
348 		gotoerr(EAFNOSUPPORT);
349 	}
350 
351 	if (dst->sa_family == pseudo_AF_HDRCMPLT)	/* unlikely */
352 		memcpy(eh->ether_shost,
353 		       ((struct ether_header *)dst->sa_data)->ether_shost,
354 		       ETHER_ADDR_LEN);
355 	else
356 		memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
357 
358 	/*
359 	 * Bridges require special output handling.
360 	 */
361 	if (ifp->if_bridge) {
362 		KASSERT(bridge_output_p != NULL,
363 			("%s: if_bridge not loaded!", __func__));
364 		return bridge_output_p(ifp, m);
365 	}
366 
367 	/*
368 	 * If a simplex interface, and the packet is being sent to our
369 	 * Ethernet address or a broadcast address, loopback a copy.
370 	 * XXX To make a simplex device behave exactly like a duplex
371 	 * device, we should copy in the case of sending to our own
372 	 * ethernet address (thus letting the original actually appear
373 	 * on the wire). However, we don't do that here for security
374 	 * reasons and compatibility with the original behavior.
375 	 */
376 	if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
377 		int csum_flags = 0;
378 
379 		if (m->m_pkthdr.csum_flags & CSUM_IP)
380 			csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
381 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
382 			csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
383 		if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
384 			struct mbuf *n;
385 
386 			if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) {
387 				n->m_pkthdr.csum_flags |= csum_flags;
388 				if (csum_flags & CSUM_DATA_VALID)
389 					n->m_pkthdr.csum_data = 0xffff;
390 				if_simloop(ifp, n, dst->sa_family, hlen);
391 			} else
392 				ifp->if_iqdrops++;
393 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
394 				ETHER_ADDR_LEN) == 0) {
395 			m->m_pkthdr.csum_flags |= csum_flags;
396 			if (csum_flags & CSUM_DATA_VALID)
397 				m->m_pkthdr.csum_data = 0xffff;
398 			if_simloop(ifp, m, dst->sa_family, hlen);
399 			return (0);	/* XXX */
400 		}
401 	}
402 
403 #ifdef CARP
404 	if (ifp->if_carp && (error = carp_output(ifp, m, dst, NULL)))
405 		goto bad;
406 #endif
407 
408 
409 	/* Handle ng_ether(4) processing, if any */
410 	if (ng_ether_output_p != NULL) {
411 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0)
412 			goto bad;
413 		if (m == NULL)
414 			return (0);
415 	}
416 
417 	/* Continue with link-layer output */
418 	return ether_output_frame(ifp, m);
419 
420 bad:
421 	m_freem(m);
422 	return (error);
423 }
424 
425 /*
426  * Ethernet link layer output routine to send a raw frame to the device.
427  *
428  * This assumes that the 14 byte Ethernet header is present and contiguous
429  * in the first mbuf.
430  */
431 int
432 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
433 {
434 	struct ip_fw *rule = NULL;
435 	int error = 0;
436 	struct altq_pktattr pktattr;
437 	struct m_tag *mtag;
438 
439 	ASSERT_NOT_SERIALIZED(ifp->if_serializer);
440 
441 	/* Extract info from dummynet tag */
442 	mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
443 	if (mtag != NULL) {
444 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
445 
446 		m_tag_delete(m, mtag);
447 		mtag = NULL;
448 	}
449 
450 	if (ifq_is_enabled(&ifp->if_snd))
451 		altq_etherclassify(&ifp->if_snd, m, &pktattr);
452 	crit_enter();
453 	if (IPFW_LOADED && ether_ipfw != 0) {
454 		struct ether_header save_eh, *eh;
455 
456 		eh = mtod(m, struct ether_header *);
457 		save_eh = *eh;
458 		m_adj(m, ETHER_HDR_LEN);
459 		if (!ether_ipfw_chk(&m, ifp, &rule, eh)) {
460 			crit_exit();
461 			if (m != NULL) {
462 				m_freem(m);
463 				return ENOBUFS; /* pkt dropped */
464 			} else
465 				return 0;	/* consumed e.g. in a pipe */
466 		}
467 
468 		/* packet was ok, restore the ethernet header */
469 		ether_restore_header(&m, eh, &save_eh);
470 		if (m == NULL) {
471 			crit_exit();
472 			return ENOBUFS;
473 		}
474 	}
475 	crit_exit();
476 
477 	/*
478 	 * Queue message on interface, update output statistics if
479 	 * successful, and start output if interface not yet active.
480 	 */
481 	error = ifq_dispatch(ifp, m, &pktattr);
482 	return (error);
483 }
484 
485 /*
486  * ipfw processing for ethernet packets (in and out).
487  * The second parameter is NULL from ether_demux(), and ifp from
488  * ether_output_frame().
489  */
490 static boolean_t
491 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule,
492 	       const struct ether_header *eh)
493 {
494 	struct ether_header save_eh = *eh;	/* might be a ptr in m */
495 	struct ip_fw_args args;
496 	struct m_tag *mtag;
497 	int i;
498 
499 	if (*rule != NULL && fw_one_pass)
500 		return TRUE; /* dummynet packet, already partially processed */
501 
502 	/*
503 	 * I need some amount of data to be contiguous.
504 	 */
505 	i = min((*m0)->m_pkthdr.len, max_protohdr);
506 	if ((*m0)->m_len < i) {
507 		*m0 = m_pullup(*m0, i);
508 		if (*m0 == NULL)
509 			return FALSE;
510 	}
511 
512 	args.m = *m0;		/* the packet we are looking at		*/
513 	args.oif = dst;		/* destination, if any			*/
514 	if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
515 		m_tag_delete(*m0, mtag);
516 	args.rule = *rule;	/* matching rule to restart		*/
517 	args.next_hop = NULL;	/* we do not support forward yet	*/
518 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
519 	i = ip_fw_chk_ptr(&args);
520 	*m0 = args.m;
521 	*rule = args.rule;
522 
523 	if ((i & IP_FW_PORT_DENY_FLAG) || *m0 == NULL)	/* drop */
524 		return FALSE;
525 
526 	if (i == 0)					/* a PASS rule.  */
527 		return TRUE;
528 
529 	if (i & IP_FW_PORT_DYNT_FLAG) {
530 		/*
531 		 * Pass the pkt to dummynet, which consumes it.
532 		 */
533 		struct mbuf *m;
534 
535 		m = *m0;	/* pass the original to dummynet */
536 		*m0 = NULL;	/* and nothing back to the caller */
537 
538 		ether_restore_header(&m, eh, &save_eh);
539 		if (m == NULL)
540 			return FALSE;
541 
542 		ip_fw_dn_io_ptr(m, (i & 0xffff),
543 			dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
544 		return FALSE;
545 	}
546 	/*
547 	 * XXX at some point add support for divert/forward actions.
548 	 * If none of the above matches, we have to drop the pkt.
549 	 */
550 	return FALSE;
551 }
552 
553 /*
554  * Process a received Ethernet packet.
555  *
556  * The ethernet header is assumed to be in the mbuf so the caller
557  * MUST MAKE SURE that there are at least sizeof(struct ether_header)
558  * bytes in the first mbuf.
559  *
560  * This allows us to concentrate in one place a bunch of code which
561  * is replicated in all device drivers. Also, many functions called
562  * from ether_input() try to put the eh back into the mbuf, so we
563  * can later propagate the 'contiguous packet' interface to them.
564  *
565  * NOTA BENE: for all drivers "eh" is a pointer into the first mbuf or
566  * cluster, right before m_data. So be very careful when working on m,
567  * as you could destroy *eh !!
568  *
569  * First we perform any link layer operations, then continue to the
570  * upper layers with ether_demux().
571  */
572 void
573 ether_input_chain(struct ifnet *ifp, struct mbuf *m, struct mbuf_chain *chain)
574 {
575 	struct ether_header *eh;
576 
577 	ASSERT_SERIALIZED(ifp->if_serializer);
578 	M_ASSERTPKTHDR(m);
579 
580 	/* Discard packet if interface is not up */
581 	if (!(ifp->if_flags & IFF_UP)) {
582 		m_freem(m);
583 		return;
584 	}
585 
586 	if (m->m_len < sizeof(struct ether_header)) {
587 		/* XXX error in the caller. */
588 		m_freem(m);
589 		return;
590 	}
591 	eh = mtod(m, struct ether_header *);
592 
593 	if (ntohs(eh->ether_type) == ETHERTYPE_VLAN &&
594 	    (m->m_flags & M_VLANTAG) == 0) {
595 		/*
596 		 * Extract vlan tag if hardware does not do it for us
597 		 */
598 		vlan_ether_decap(&m);
599 		if (m == NULL)
600 			return;
601 		eh = mtod(m, struct ether_header *);
602 	}
603 
604 	m->m_pkthdr.rcvif = ifp;
605 
606 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
607 		if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
608 			 ifp->if_addrlen) == 0)
609 			m->m_flags |= M_BCAST;
610 		else
611 			m->m_flags |= M_MCAST;
612 		ifp->if_imcasts++;
613 	}
614 
615 	ETHER_BPF_MTAP(ifp, m);
616 
617 	ifp->if_ibytes += m->m_pkthdr.len;
618 
619 	if (ifp->if_flags & IFF_MONITOR) {
620 		/*
621 		 * Interface marked for monitoring; discard packet.
622 		 */
623 		 m_freem(m);
624 		 return;
625 	}
626 
627 	/*
628 	 * Tap the packet off here for a bridge.  bridge_input()
629 	 * will return NULL if it has consumed the packet, otherwise
630 	 * it gets processed as normal.  Note that bridge_input()
631 	 * will always return the original packet if we need to
632 	 * process it locally.
633 	 */
634 	if (ifp->if_bridge) {
635 		KASSERT(bridge_input_p != NULL,
636 			("%s: if_bridge not loaded!", __func__));
637 
638 		if(m->m_flags & M_PROTO1) {
639 			m->m_flags &= ~M_PROTO1;
640 		} else {
641 			/* clear M_PROMISC, in case the packets comes from a vlan */
642 			/* m->m_flags &= ~M_PROMISC; */
643 			lwkt_serialize_exit(ifp->if_serializer);
644 			m = bridge_input_p(ifp, m);
645 			lwkt_serialize_enter(ifp->if_serializer);
646 			if (m == NULL)
647 				return;
648 
649 			KASSERT(ifp == m->m_pkthdr.rcvif,
650 				("bridge_input_p changed rcvif\n"));
651 
652 			/* 'm' may be changed by bridge_input_p() */
653 			eh = mtod(m, struct ether_header *);
654 		}
655 	}
656 
657 	/* Handle ng_ether(4) processing, if any */
658 	if (ng_ether_input_p != NULL) {
659 		ng_ether_input_p(ifp, &m);
660 		if (m == NULL)
661 			return;
662 
663 		/* 'm' may be changed by ng_ether_input_p() */
664 		eh = mtod(m, struct ether_header *);
665 	}
666 
667 	/* Continue with upper layer processing */
668 	ether_demux_chain(ifp, m, chain);
669 }
670 
671 void
672 ether_input(struct ifnet *ifp, struct mbuf *m)
673 {
674 	ether_input_chain(ifp, m, NULL);
675 }
676 
677 /*
678  * Upper layer processing for a received Ethernet packet.
679  */
680 static void
681 ether_demux_chain(struct ifnet *ifp, struct mbuf *m, struct mbuf_chain *chain)
682 {
683 	struct ether_header save_eh, *eh;
684 	int isr;
685 	u_short ether_type;
686 	struct ip_fw *rule = NULL;
687 	struct m_tag *mtag;
688 #ifdef NETATALK
689 	struct llc *l;
690 #endif
691 
692 	M_ASSERTPKTHDR(m);
693 	KASSERT(m->m_len >= ETHER_HDR_LEN,
694 		("ether header is no contiguous!\n"));
695 
696 	eh = mtod(m, struct ether_header *);
697 	save_eh = *eh;
698 
699 	/* XXX old crufty stuff, needs to be removed */
700 	m_adj(m, sizeof(struct ether_header));
701 
702 	/* Extract info from dummynet tag */
703 	mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
704 	if (mtag != NULL) {
705 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
706 		KKASSERT(ifp == NULL);
707 		ifp = m->m_pkthdr.rcvif;
708 
709 		m_tag_delete(m, mtag);
710 		mtag = NULL;
711 	}
712 	if (rule)	/* packet is passing the second time */
713 		goto post_stats;
714 
715 #ifdef CARP
716 	/*
717 	 * XXX: Okay, we need to call carp_forus() and - if it is for
718 	 * us jump over code that does the normal check
719 	 * "ac_enaddr == ether_dhost". The check sequence is a bit
720 	 * different from OpenBSD, so we jump over as few code as
721 	 * possible, to catch _all_ sanity checks. This needs
722 	 * evaluation, to see if the carp ether_dhost values break any
723 	 * of these checks!
724 	 */
725 	if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost))
726 		goto post_stats;
727 #endif
728 
729 	/*
730 	 * Discard packet if upper layers shouldn't see it because
731 	 * it was unicast to a different Ethernet address.  If the
732 	 * driver is working properly, then this situation can only
733 	 * happen when the interface is in promiscuous mode.
734 	 */
735 	if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
736 	    (eh->ether_dhost[0] & 1) == 0 &&
737 	    bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
738 		m_freem(m);
739 		return;
740 	}
741 
742 post_stats:
743 	if (IPFW_LOADED && ether_ipfw != 0) {
744 		if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
745 			m_freem(m);
746 			return;
747 		}
748 	}
749 
750 	ether_type = ntohs(save_eh.ether_type);
751 
752 	if (m->m_flags & M_VLANTAG) {
753 		if (ether_type == ETHERTYPE_VLAN) {
754 			/*
755 			 * To prevent possible dangerous recursion,
756 			 * we don't do vlan-in-vlan
757 			 */
758 			m->m_pkthdr.rcvif->if_noproto++;
759 			m_freem(m);
760 			return;
761 		}
762 
763 		if (vlan_input_p != NULL) {
764 			ether_restore_header(&m, eh, &save_eh);
765 			if (m != NULL)
766 				vlan_input_p(m, chain);
767 		} else {
768 			m->m_pkthdr.rcvif->if_noproto++;
769 			m_freem(m);
770 		}
771 		return;
772 	}
773 	KKASSERT(ether_type != ETHERTYPE_VLAN);
774 
775 	switch (ether_type) {
776 #ifdef INET
777 	case ETHERTYPE_IP:
778 		if (ipflow_fastforward(m, ifp->if_serializer))
779 			return;
780 		isr = NETISR_IP;
781 		break;
782 
783 	case ETHERTYPE_ARP:
784 		if (ifp->if_flags & IFF_NOARP) {
785 			/* Discard packet if ARP is disabled on interface */
786 			m_freem(m);
787 			return;
788 		}
789 		isr = NETISR_ARP;
790 		break;
791 #endif
792 
793 #ifdef INET6
794 	case ETHERTYPE_IPV6:
795 		isr = NETISR_IPV6;
796 		break;
797 #endif
798 
799 #ifdef IPX
800 	case ETHERTYPE_IPX:
801 		if (ef_inputp && ef_inputp(ifp, &save_eh, m) == 0)
802 			return;
803 		isr = NETISR_IPX;
804 		break;
805 #endif
806 
807 #ifdef NS
808 	case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
809 		isr = NETISR_NS;
810 		break;
811 
812 #endif
813 
814 #ifdef NETATALK
815 	case ETHERTYPE_AT:
816 		isr = NETISR_ATALK1;
817 		break;
818 	case ETHERTYPE_AARP:
819 		isr = NETISR_AARP;
820 		break;
821 #endif
822 
823 #ifdef MPLS
824 	case ETHERTYPE_MPLS:
825 	case ETHERTYPE_MPLS_MCAST:
826 		isr = NETISR_MPLS;
827 		break;
828 #endif
829 
830 	default:
831 #ifdef IPX
832 		if (ef_inputp && ef_inputp(ifp, &save_eh, m) == 0)
833 			return;
834 #endif
835 #ifdef NS
836 		checksum = mtod(m, ushort *);
837 		/* Novell 802.3 */
838 		if ((ether_type <= ETHERMTU) &&
839 		    ((*checksum == 0xffff) || (*checksum == 0xE0E0))) {
840 			if (*checksum == 0xE0E0) {
841 				m->m_pkthdr.len -= 3;
842 				m->m_len -= 3;
843 				m->m_data += 3;
844 			}
845 			isr = NETISR_NS;
846 			break;
847 		}
848 #endif
849 #ifdef NETATALK
850 		if (ether_type > ETHERMTU)
851 			goto dropanyway;
852 		l = mtod(m, struct llc *);
853 		if (l->llc_dsap == LLC_SNAP_LSAP &&
854 		    l->llc_ssap == LLC_SNAP_LSAP &&
855 		    l->llc_control == LLC_UI) {
856 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
857 				 sizeof at_org_code) == 0 &&
858 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
859 				m_adj(m, sizeof(struct llc));
860 				isr = NETISR_ATALK2;
861 				break;
862 			}
863 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
864 				 sizeof aarp_org_code) == 0 &&
865 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
866 				m_adj(m, sizeof(struct llc));
867 				isr = NETISR_AARP;
868 				break;
869 			}
870 		}
871 dropanyway:
872 #endif
873 		if (ng_ether_input_orphan_p != NULL)
874 			(*ng_ether_input_orphan_p)(ifp, m, &save_eh);
875 		else
876 			m_freem(m);
877 		return;
878 	}
879 
880 #ifdef ETHER_INPUT_CHAIN
881 	if (chain != NULL) {
882 		struct mbuf_chain *c;
883 		lwkt_port_t port;
884 		int cpuid;
885 
886 		port = netisr_mport(isr, &m);
887 		if (port == NULL)
888 			return;
889 
890 		m->m_pkthdr.header = port; /* XXX */
891 		cpuid = port->mpu_td->td_gd->gd_cpuid;
892 
893 		c = &chain[cpuid];
894 		if (c->mc_head == NULL) {
895 			c->mc_head = c->mc_tail = m;
896 		} else {
897 			c->mc_tail->m_nextpkt = m;
898 			c->mc_tail = m;
899 		}
900 		m->m_nextpkt = NULL;
901 	} else
902 #endif	/* ETHER_INPUT_CHAIN */
903 		netisr_dispatch(isr, m);
904 }
905 
906 void
907 ether_demux(struct ifnet *ifp, struct mbuf *m)
908 {
909 	ether_demux_chain(ifp, m, NULL);
910 }
911 
912 /*
913  * Perform common duties while attaching to interface list
914  */
915 
916 void
917 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer)
918 {
919 	ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
920 			   serializer);
921 }
922 
923 void
924 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen,
925 		   lwkt_serialize_t serializer)
926 {
927 	struct sockaddr_dl *sdl;
928 
929 	ifp->if_type = IFT_ETHER;
930 	ifp->if_addrlen = ETHER_ADDR_LEN;
931 	ifp->if_hdrlen = ETHER_HDR_LEN;
932 	if_attach(ifp, serializer);
933 	ifp->if_mtu = ETHERMTU;
934 	if (ifp->if_baudrate == 0)
935 		ifp->if_baudrate = 10000000;
936 	ifp->if_output = ether_output;
937 	ifp->if_input = ether_input;
938 	ifp->if_resolvemulti = ether_resolvemulti;
939 	ifp->if_broadcastaddr = etherbroadcastaddr;
940 	sdl = IF_LLSOCKADDR(ifp);
941 	sdl->sdl_type = IFT_ETHER;
942 	sdl->sdl_alen = ifp->if_addrlen;
943 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
944 	/*
945 	 * XXX Keep the current drivers happy.
946 	 * XXX Remove once all drivers have been cleaned up
947 	 */
948 	if (lla != IFP2AC(ifp)->ac_enaddr)
949 		bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
950 	bpfattach(ifp, dlt, hdrlen);
951 	if (ng_ether_attach_p != NULL)
952 		(*ng_ether_attach_p)(ifp);
953 
954 	if_printf(ifp, "MAC address: %6D\n", lla, ":");
955 }
956 
957 /*
958  * Perform common duties while detaching an Ethernet interface
959  */
960 void
961 ether_ifdetach(struct ifnet *ifp)
962 {
963 	if_down(ifp);
964 
965 	if (ng_ether_detach_p != NULL)
966 		(*ng_ether_detach_p)(ifp);
967 	bpfdetach(ifp);
968 	if_detach(ifp);
969 }
970 
971 int
972 ether_ioctl(struct ifnet *ifp, int command, caddr_t data)
973 {
974 	struct ifaddr *ifa = (struct ifaddr *) data;
975 	struct ifreq *ifr = (struct ifreq *) data;
976 	int error = 0;
977 
978 #define IF_INIT(ifp) \
979 do { \
980 	if (((ifp)->if_flags & IFF_UP) == 0) { \
981 		(ifp)->if_flags |= IFF_UP; \
982 		(ifp)->if_init((ifp)->if_softc); \
983 	} \
984 } while (0)
985 
986 	ASSERT_SERIALIZED(ifp->if_serializer);
987 
988 	switch (command) {
989 	case SIOCSIFADDR:
990 		switch (ifa->ifa_addr->sa_family) {
991 #ifdef INET
992 		case AF_INET:
993 			IF_INIT(ifp);	/* before arpwhohas */
994 			arp_ifinit(ifp, ifa);
995 			break;
996 #endif
997 #ifdef IPX
998 		/*
999 		 * XXX - This code is probably wrong
1000 		 */
1001 		case AF_IPX:
1002 			{
1003 			struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr;
1004 			struct arpcom *ac = IFP2AC(ifp);
1005 
1006 			if (ipx_nullhost(*ina))
1007 				ina->x_host = *(union ipx_host *) ac->ac_enaddr;
1008 			else
1009 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
1010 				      sizeof ac->ac_enaddr);
1011 
1012 			IF_INIT(ifp);	/* Set new address. */
1013 			break;
1014 			}
1015 #endif
1016 #ifdef NS
1017 		/*
1018 		 * XXX - This code is probably wrong
1019 		 */
1020 		case AF_NS:
1021 		{
1022 			struct ns_addr *ina = &(IA_SNS(ifa)->sns_addr);
1023 			struct arpcom *ac = IFP2AC(ifp);
1024 
1025 			if (ns_nullhost(*ina))
1026 				ina->x_host = *(union ns_host *)(ac->ac_enaddr);
1027 			else
1028 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
1029 				      sizeof ac->ac_enaddr);
1030 
1031 			/*
1032 			 * Set new address
1033 			 */
1034 			IF_INIT(ifp);
1035 			break;
1036 		}
1037 #endif
1038 		default:
1039 			IF_INIT(ifp);
1040 			break;
1041 		}
1042 		break;
1043 
1044 	case SIOCGIFADDR:
1045 		bcopy(IFP2AC(ifp)->ac_enaddr,
1046 		      ((struct sockaddr *)ifr->ifr_data)->sa_data,
1047 		      ETHER_ADDR_LEN);
1048 		break;
1049 
1050 	case SIOCSIFMTU:
1051 		/*
1052 		 * Set the interface MTU.
1053 		 */
1054 		if (ifr->ifr_mtu > ETHERMTU) {
1055 			error = EINVAL;
1056 		} else {
1057 			ifp->if_mtu = ifr->ifr_mtu;
1058 		}
1059 		break;
1060 	default:
1061 		error = EINVAL;
1062 		break;
1063 	}
1064 	return (error);
1065 
1066 #undef IF_INIT
1067 }
1068 
1069 int
1070 ether_resolvemulti(
1071 	struct ifnet *ifp,
1072 	struct sockaddr **llsa,
1073 	struct sockaddr *sa)
1074 {
1075 	struct sockaddr_dl *sdl;
1076 	struct sockaddr_in *sin;
1077 #ifdef INET6
1078 	struct sockaddr_in6 *sin6;
1079 #endif
1080 	u_char *e_addr;
1081 
1082 	switch(sa->sa_family) {
1083 	case AF_LINK:
1084 		/*
1085 		 * No mapping needed. Just check that it's a valid MC address.
1086 		 */
1087 		sdl = (struct sockaddr_dl *)sa;
1088 		e_addr = LLADDR(sdl);
1089 		if ((e_addr[0] & 1) != 1)
1090 			return EADDRNOTAVAIL;
1091 		*llsa = 0;
1092 		return 0;
1093 
1094 #ifdef INET
1095 	case AF_INET:
1096 		sin = (struct sockaddr_in *)sa;
1097 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1098 			return EADDRNOTAVAIL;
1099 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
1100 		       M_WAITOK | M_ZERO);
1101 		sdl->sdl_len = sizeof *sdl;
1102 		sdl->sdl_family = AF_LINK;
1103 		sdl->sdl_index = ifp->if_index;
1104 		sdl->sdl_type = IFT_ETHER;
1105 		sdl->sdl_alen = ETHER_ADDR_LEN;
1106 		e_addr = LLADDR(sdl);
1107 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1108 		*llsa = (struct sockaddr *)sdl;
1109 		return 0;
1110 #endif
1111 #ifdef INET6
1112 	case AF_INET6:
1113 		sin6 = (struct sockaddr_in6 *)sa;
1114 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1115 			/*
1116 			 * An IP6 address of 0 means listen to all
1117 			 * of the Ethernet multicast address used for IP6.
1118 			 * (This is used for multicast routers.)
1119 			 */
1120 			ifp->if_flags |= IFF_ALLMULTI;
1121 			*llsa = 0;
1122 			return 0;
1123 		}
1124 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1125 			return EADDRNOTAVAIL;
1126 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
1127 		       M_WAITOK | M_ZERO);
1128 		sdl->sdl_len = sizeof *sdl;
1129 		sdl->sdl_family = AF_LINK;
1130 		sdl->sdl_index = ifp->if_index;
1131 		sdl->sdl_type = IFT_ETHER;
1132 		sdl->sdl_alen = ETHER_ADDR_LEN;
1133 		e_addr = LLADDR(sdl);
1134 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1135 		*llsa = (struct sockaddr *)sdl;
1136 		return 0;
1137 #endif
1138 
1139 	default:
1140 		/*
1141 		 * Well, the text isn't quite right, but it's the name
1142 		 * that counts...
1143 		 */
1144 		return EAFNOSUPPORT;
1145 	}
1146 }
1147 
1148 #if 0
1149 /*
1150  * This is for reference.  We have a table-driven version
1151  * of the little-endian crc32 generator, which is faster
1152  * than the double-loop.
1153  */
1154 uint32_t
1155 ether_crc32_le(const uint8_t *buf, size_t len)
1156 {
1157 	uint32_t c, crc, carry;
1158 	size_t i, j;
1159 
1160 	crc = 0xffffffffU;	/* initial value */
1161 
1162 	for (i = 0; i < len; i++) {
1163 		c = buf[i];
1164 		for (j = 0; j < 8; j++) {
1165 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1166 			crc >>= 1;
1167 			c >>= 1;
1168 			if (carry)
1169 				crc = (crc ^ ETHER_CRC_POLY_LE);
1170 		}
1171 	}
1172 
1173 	return (crc);
1174 }
1175 #else
1176 uint32_t
1177 ether_crc32_le(const uint8_t *buf, size_t len)
1178 {
1179 	static const uint32_t crctab[] = {
1180 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1181 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1182 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1183 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1184 	};
1185 	uint32_t crc;
1186 	size_t i;
1187 
1188 	crc = 0xffffffffU;	/* initial value */
1189 
1190 	for (i = 0; i < len; i++) {
1191 		crc ^= buf[i];
1192 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1193 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1194 	}
1195 
1196 	return (crc);
1197 }
1198 #endif
1199 
1200 uint32_t
1201 ether_crc32_be(const uint8_t *buf, size_t len)
1202 {
1203 	uint32_t c, crc, carry;
1204 	size_t i, j;
1205 
1206 	crc = 0xffffffffU;	/* initial value */
1207 
1208 	for (i = 0; i < len; i++) {
1209 		c = buf[i];
1210 		for (j = 0; j < 8; j++) {
1211 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1212 			crc <<= 1;
1213 			c >>= 1;
1214 			if (carry)
1215 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1216 		}
1217 	}
1218 
1219 	return (crc);
1220 }
1221 
1222 /*
1223  * find the size of ethernet header, and call classifier
1224  */
1225 void
1226 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
1227 		   struct altq_pktattr *pktattr)
1228 {
1229 	struct ether_header *eh;
1230 	uint16_t ether_type;
1231 	int hlen, af, hdrsize;
1232 	caddr_t hdr;
1233 
1234 	hlen = sizeof(struct ether_header);
1235 	eh = mtod(m, struct ether_header *);
1236 
1237 	ether_type = ntohs(eh->ether_type);
1238 	if (ether_type < ETHERMTU) {
1239 		/* ick! LLC/SNAP */
1240 		struct llc *llc = (struct llc *)(eh + 1);
1241 		hlen += 8;
1242 
1243 		if (m->m_len < hlen ||
1244 		    llc->llc_dsap != LLC_SNAP_LSAP ||
1245 		    llc->llc_ssap != LLC_SNAP_LSAP ||
1246 		    llc->llc_control != LLC_UI)
1247 			goto bad;  /* not snap! */
1248 
1249 		ether_type = ntohs(llc->llc_un.type_snap.ether_type);
1250 	}
1251 
1252 	if (ether_type == ETHERTYPE_IP) {
1253 		af = AF_INET;
1254 		hdrsize = 20;  /* sizeof(struct ip) */
1255 #ifdef INET6
1256 	} else if (ether_type == ETHERTYPE_IPV6) {
1257 		af = AF_INET6;
1258 		hdrsize = 40;  /* sizeof(struct ip6_hdr) */
1259 #endif
1260 	} else
1261 		goto bad;
1262 
1263 	while (m->m_len <= hlen) {
1264 		hlen -= m->m_len;
1265 		m = m->m_next;
1266 	}
1267 	hdr = m->m_data + hlen;
1268 	if (m->m_len < hlen + hdrsize) {
1269 		/*
1270 		 * ip header is not in a single mbuf.  this should not
1271 		 * happen in the current code.
1272 		 * (todo: use m_pulldown in the future)
1273 		 */
1274 		goto bad;
1275 	}
1276 	m->m_data += hlen;
1277 	m->m_len -= hlen;
1278 	ifq_classify(ifq, m, af, pktattr);
1279 	m->m_data -= hlen;
1280 	m->m_len += hlen;
1281 
1282 	return;
1283 
1284 bad:
1285 	pktattr->pattr_class = NULL;
1286 	pktattr->pattr_hdr = NULL;
1287 	pktattr->pattr_af = AF_UNSPEC;
1288 }
1289 
1290 static void
1291 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
1292 		     const struct ether_header *save_eh)
1293 {
1294 	struct mbuf *m = *m0;
1295 
1296 	ether_restore_hdr++;
1297 
1298 	/*
1299 	 * Prepend the header, optimize for the common case of
1300 	 * eh pointing into the mbuf.
1301 	 */
1302 	if ((const void *)(eh + 1) == (void *)m->m_data) {
1303 		m->m_data -= ETHER_HDR_LEN;
1304 		m->m_len += ETHER_HDR_LEN;
1305 		m->m_pkthdr.len += ETHER_HDR_LEN;
1306 	} else {
1307 		ether_prepend_hdr++;
1308 
1309 		M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
1310 		if (m != NULL) {
1311 			bcopy(save_eh, mtod(m, struct ether_header *),
1312 			      ETHER_HDR_LEN);
1313 		}
1314 	}
1315 	*m0 = m;
1316 }
1317 
1318 #ifdef ETHER_INPUT_CHAIN
1319 
1320 static void
1321 ether_input_ipifunc(void *arg)
1322 {
1323 	struct mbuf *m, *next;
1324 	lwkt_port_t port;
1325 
1326 	m = arg;
1327 	do {
1328 		next = m->m_nextpkt;
1329 		m->m_nextpkt = NULL;
1330 
1331 		port = m->m_pkthdr.header;
1332 		m->m_pkthdr.header = NULL;
1333 
1334 		lwkt_sendmsg(port,
1335 		&m->m_hdr.mh_netmsg.nm_netmsg.nm_lmsg);
1336 
1337 		m = next;
1338 	} while (m != NULL);
1339 }
1340 
1341 void
1342 ether_input_dispatch(struct mbuf_chain *chain)
1343 {
1344 #ifdef SMP
1345 	int i;
1346 
1347 	for (i = 0; i < ncpus; ++i) {
1348 		if (chain[i].mc_head != NULL) {
1349 			lwkt_send_ipiq(globaldata_find(i),
1350 			ether_input_ipifunc, chain[i].mc_head);
1351 		}
1352 	}
1353 #else
1354 	if (chain->mc_head != NULL)
1355 		ether_input_ipifunc(chain->mc_head);
1356 #endif
1357 }
1358 
1359 void
1360 ether_input_chain_init(struct mbuf_chain *chain)
1361 {
1362 #ifdef SMP
1363 	int i;
1364 
1365 	for (i = 0; i < ncpus; ++i)
1366 		chain[i].mc_head = chain[i].mc_tail = NULL;
1367 #else
1368 	chain->mc_head = chain->mc_tail = NULL;
1369 #endif
1370 }
1371 
1372 #endif	/* ETHER_INPUT_CHAIN */
1373 
1374 #ifdef ETHER_INPUT2
1375 
1376 static void
1377 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
1378 {
1379 	struct ether_header *eh;
1380 	int isr, redispatch;
1381 	u_short ether_type;
1382 	struct ip_fw *rule = NULL;
1383 	struct m_tag *mtag;
1384 #ifdef NETATALK
1385 	struct llc *l;
1386 #endif
1387 
1388 	M_ASSERTPKTHDR(m);
1389 	KASSERT(m->m_len >= ETHER_HDR_LEN,
1390 		("ether header is no contiguous!\n"));
1391 
1392 	eh = mtod(m, struct ether_header *);
1393 
1394 	/* Extract info from dummynet tag */
1395 	mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
1396 	if (mtag != NULL) {
1397 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
1398 		KKASSERT(ifp == NULL);
1399 		ifp = m->m_pkthdr.rcvif;
1400 
1401 		m_tag_delete(m, mtag);
1402 		mtag = NULL;
1403 	}
1404 	if (rule)	/* packet is passing the second time */
1405 		goto post_stats;
1406 
1407 #ifdef CARP
1408 	/*
1409 	 * XXX: Okay, we need to call carp_forus() and - if it is for
1410 	 * us jump over code that does the normal check
1411 	 * "ac_enaddr == ether_dhost". The check sequence is a bit
1412 	 * different from OpenBSD, so we jump over as few code as
1413 	 * possible, to catch _all_ sanity checks. This needs
1414 	 * evaluation, to see if the carp ether_dhost values break any
1415 	 * of these checks!
1416 	 */
1417 	if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost))
1418 		goto post_stats;
1419 #endif
1420 
1421 	/*
1422 	 * Discard packet if upper layers shouldn't see it because
1423 	 * it was unicast to a different Ethernet address.  If the
1424 	 * driver is working properly, then this situation can only
1425 	 * happen when the interface is in promiscuous mode.
1426 	 */
1427 	if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
1428 	    (eh->ether_dhost[0] & 1) == 0 &&
1429 	    bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
1430 		m_freem(m);
1431 		return;
1432 	}
1433 
1434 post_stats:
1435 	if (IPFW_LOADED && ether_ipfw != 0) {
1436 		struct ether_header save_eh = *eh;
1437 
1438 		/* XXX old crufty stuff, needs to be removed */
1439 		m_adj(m, sizeof(struct ether_header));
1440 
1441 		if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
1442 			m_freem(m);
1443 			return;
1444 		}
1445 
1446 		ether_restore_header(&m, eh, &save_eh);
1447 		if (m == NULL)
1448 			return;
1449 		eh = mtod(m, struct ether_header *);
1450 	}
1451 
1452 	ether_type = ntohs(eh->ether_type);
1453 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1454 
1455 	if (m->m_flags & M_VLANTAG) {
1456 		if (vlan_input2_p != NULL) {
1457 			vlan_input2_p(m);
1458 		} else {
1459 			m->m_pkthdr.rcvif->if_noproto++;
1460 			m_freem(m);
1461 		}
1462 		return;
1463 	}
1464 
1465 	m_adj(m, sizeof(struct ether_header));
1466 	redispatch = 0;
1467 
1468 	switch (ether_type) {
1469 #ifdef INET
1470 	case ETHERTYPE_IP:
1471 #ifdef notyet
1472 		if (ipflow_fastforward(m, ifp->if_serializer))
1473 			return;
1474 #endif
1475 		isr = NETISR_IP;
1476 		break;
1477 
1478 	case ETHERTYPE_ARP:
1479 		if (ifp->if_flags & IFF_NOARP) {
1480 			/* Discard packet if ARP is disabled on interface */
1481 			m_freem(m);
1482 			return;
1483 		}
1484 		isr = NETISR_ARP;
1485 		break;
1486 #endif
1487 
1488 #ifdef INET6
1489 	case ETHERTYPE_IPV6:
1490 		isr = NETISR_IPV6;
1491 		break;
1492 #endif
1493 
1494 #ifdef IPX
1495 	case ETHERTYPE_IPX:
1496 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
1497 			return;
1498 		isr = NETISR_IPX;
1499 		break;
1500 #endif
1501 
1502 #ifdef NS
1503 	case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
1504 		isr = NETISR_NS;
1505 		break;
1506 
1507 #endif
1508 
1509 #ifdef NETATALK
1510 	case ETHERTYPE_AT:
1511 		isr = NETISR_ATALK1;
1512 		break;
1513 	case ETHERTYPE_AARP:
1514 		isr = NETISR_AARP;
1515 		break;
1516 #endif
1517 
1518 	default:
1519 		/*
1520 		 * The accurate msgport is not determined before
1521 		 * we reach here, so redo the dispatching
1522 		 */
1523 		redispatch = 1;
1524 #ifdef IPX
1525 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
1526 			return;
1527 #endif
1528 #ifdef NS
1529 		checksum = mtod(m, ushort *);
1530 		/* Novell 802.3 */
1531 		if ((ether_type <= ETHERMTU) &&
1532 		    ((*checksum == 0xffff) || (*checksum == 0xE0E0))) {
1533 			if (*checksum == 0xE0E0) {
1534 				m->m_pkthdr.len -= 3;
1535 				m->m_len -= 3;
1536 				m->m_data += 3;
1537 			}
1538 			isr = NETISR_NS;
1539 			break;
1540 		}
1541 #endif
1542 #ifdef NETATALK
1543 		if (ether_type > ETHERMTU)
1544 			goto dropanyway;
1545 		l = mtod(m, struct llc *);
1546 		if (l->llc_dsap == LLC_SNAP_LSAP &&
1547 		    l->llc_ssap == LLC_SNAP_LSAP &&
1548 		    l->llc_control == LLC_UI) {
1549 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
1550 				 sizeof at_org_code) == 0 &&
1551 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
1552 				m_adj(m, sizeof(struct llc));
1553 				isr = NETISR_ATALK2;
1554 				break;
1555 			}
1556 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
1557 				 sizeof aarp_org_code) == 0 &&
1558 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
1559 				m_adj(m, sizeof(struct llc));
1560 				isr = NETISR_AARP;
1561 				break;
1562 			}
1563 		}
1564 dropanyway:
1565 #endif
1566 		if (ng_ether_input_orphan_p != NULL)
1567 			ng_ether_input_orphan_p(ifp, m, eh);
1568 		else
1569 			m_freem(m);
1570 		return;
1571 	}
1572 
1573 	if (!redispatch)
1574 		netisr_run(isr, m);
1575 	else
1576 		netisr_dispatch(isr, m);
1577 }
1578 
1579 void
1580 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1581 {
1582 	if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1583 		/*
1584 		 * Receiving interface's flags are changed, when this
1585 		 * packet is waiting for processing; discard it.
1586 		 */
1587 		m_freem(m);
1588 		return;
1589 	}
1590 
1591 	/*
1592 	 * Tap the packet off here for a bridge.  bridge_input()
1593 	 * will return NULL if it has consumed the packet, otherwise
1594 	 * it gets processed as normal.  Note that bridge_input()
1595 	 * will always return the original packet if we need to
1596 	 * process it locally.
1597 	 */
1598 	if (ifp->if_bridge) {
1599 		KASSERT(bridge_input_p != NULL,
1600 			("%s: if_bridge not loaded!", __func__));
1601 
1602 		if(m->m_flags & M_PROTO1) {
1603 			m->m_flags &= ~M_PROTO1;
1604 		} else {
1605 			/* clear M_PROMISC, in case the packets comes from a vlan */
1606 			/* m->m_flags &= ~M_PROMISC; */
1607 			m = bridge_input_p(ifp, m);
1608 			if (m == NULL)
1609 				return;
1610 
1611 			KASSERT(ifp == m->m_pkthdr.rcvif,
1612 				("bridge_input_p changed rcvif\n"));
1613 		}
1614 	}
1615 
1616 	/* Handle ng_ether(4) processing, if any */
1617 	if (ng_ether_input_p != NULL) {
1618 		ng_ether_input_p(ifp, &m);
1619 		if (m == NULL)
1620 			return;
1621 	}
1622 
1623 	/* Continue with upper layer processing */
1624 	ether_demux_oncpu(ifp, m);
1625 }
1626 
1627 static void
1628 ether_input_handler(struct netmsg *nmsg)
1629 {
1630 	struct netmsg_packet *nmp = (struct netmsg_packet *)nmsg;
1631 	struct ifnet *ifp;
1632 	struct mbuf *m;
1633 
1634 	m = nmp->nm_packet;
1635 	M_ASSERTPKTHDR(m);
1636 	ifp = m->m_pkthdr.rcvif;
1637 
1638 	ether_input_oncpu(ifp, m);
1639 }
1640 
1641 static __inline void
1642 ether_init_netpacket(int num, struct mbuf *m)
1643 {
1644 	struct netmsg_packet *pmsg;
1645 
1646 	pmsg = &m->m_hdr.mh_netmsg;
1647 	netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport, 0,
1648 		    ether_input_handler);
1649 	pmsg->nm_packet = m;
1650 	pmsg->nm_netmsg.nm_lmsg.u.ms_result = num;
1651 }
1652 
1653 static __inline struct lwkt_port *
1654 ether_mport(int num, struct mbuf **m0)
1655 {
1656 	struct lwkt_port *port;
1657 	struct mbuf *m = *m0;
1658 
1659 	if (num == NETISR_MAX) {
1660 		/*
1661 		 * All packets whose target msgports can't be
1662 		 * determined here are dispatched to netisr0,
1663 		 * where further dispatching may happen.
1664 		 */
1665 		return cpu_portfn(0);
1666 	}
1667 
1668 	port = netisr_find_port(num, &m);
1669 	if (port == NULL)
1670 		return NULL;
1671 
1672 	*m0 = m;
1673 	return port;
1674 }
1675 
1676 void
1677 ether_input_chain2(struct ifnet *ifp, struct mbuf *m, struct mbuf_chain *chain)
1678 {
1679 	struct ether_header *eh, *save_eh, save_eh0;
1680 	struct lwkt_port *port;
1681 	uint16_t ether_type;
1682 	int isr;
1683 
1684 	ASSERT_SERIALIZED(ifp->if_serializer);
1685 	M_ASSERTPKTHDR(m);
1686 
1687 	/* Discard packet if interface is not up */
1688 	if (!(ifp->if_flags & IFF_UP)) {
1689 		m_freem(m);
1690 		return;
1691 	}
1692 
1693 	if (m->m_len < sizeof(struct ether_header)) {
1694 		/* XXX error in the caller. */
1695 		m_freem(m);
1696 		return;
1697 	}
1698 	eh = mtod(m, struct ether_header *);
1699 
1700 	m->m_pkthdr.rcvif = ifp;
1701 
1702 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1703 		if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1704 			 ifp->if_addrlen) == 0)
1705 			m->m_flags |= M_BCAST;
1706 		else
1707 			m->m_flags |= M_MCAST;
1708 		ifp->if_imcasts++;
1709 	}
1710 
1711 	ETHER_BPF_MTAP(ifp, m);
1712 
1713 	ifp->if_ibytes += m->m_pkthdr.len;
1714 
1715 	if (ifp->if_flags & IFF_MONITOR) {
1716 		/*
1717 		 * Interface marked for monitoring; discard packet.
1718 		 */
1719 		m_freem(m);
1720 		return;
1721 	}
1722 
1723 	if (ntohs(eh->ether_type) == ETHERTYPE_VLAN &&
1724 	    (m->m_flags & M_VLANTAG) == 0) {
1725 		/*
1726 		 * Extract vlan tag if hardware does not do it for us
1727 		 */
1728 		vlan_ether_decap(&m);
1729 		if (m == NULL)
1730 			return;
1731 		eh = mtod(m, struct ether_header *);
1732 	}
1733 	ether_type = ntohs(eh->ether_type);
1734 
1735 	if ((m->m_flags & M_VLANTAG) && ether_type == ETHERTYPE_VLAN) {
1736 		/*
1737 		 * To prevent possible dangerous recursion,
1738 		 * we don't do vlan-in-vlan
1739 		 */
1740 		ifp->if_noproto++;
1741 		m_freem(m);
1742 		return;
1743 	}
1744 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1745 
1746 	/*
1747 	 * Map ether type to netisr id.
1748 	 */
1749 	switch (ether_type) {
1750 #ifdef INET
1751 	case ETHERTYPE_IP:
1752 		isr = NETISR_IP;
1753 		break;
1754 
1755 	case ETHERTYPE_ARP:
1756 		isr = NETISR_ARP;
1757 		break;
1758 #endif
1759 
1760 #ifdef INET6
1761 	case ETHERTYPE_IPV6:
1762 		isr = NETISR_IPV6;
1763 		break;
1764 #endif
1765 
1766 #ifdef IPX
1767 	case ETHERTYPE_IPX:
1768 		isr = NETISR_IPX;
1769 		break;
1770 #endif
1771 
1772 #ifdef NS
1773 	case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
1774 		isr = NETISR_NS;
1775 		break;
1776 #endif
1777 
1778 #ifdef NETATALK
1779 	case ETHERTYPE_AT:
1780 		isr = NETISR_ATALK1;
1781 		break;
1782 	case ETHERTYPE_AARP:
1783 		isr = NETISR_AARP;
1784 		break;
1785 #endif
1786 
1787 	default:
1788 		/*
1789 		 * NETISR_MAX is an invalid value; it is chosen to let
1790 		 * ether_mport() know that we are not able to decide
1791 		 * this packet's msgport here.
1792 		 */
1793 		isr = NETISR_MAX;
1794 		break;
1795 	}
1796 
1797 	/*
1798 	 * If the packet is in contiguous memory, following
1799 	 * m_adj() could ensure that the hidden ether header
1800 	 * will not be destroyed, else we will have to save
1801 	 * the ether header for the later restoration.
1802 	 */
1803 	if (m->m_pkthdr.len != m->m_len) {
1804 		save_eh0 = *eh;
1805 		save_eh = &save_eh0;
1806 	} else {
1807 		save_eh = NULL;
1808 	}
1809 
1810 	/*
1811 	 * Temporarily remove ether header; ether_mport()
1812 	 * expects a packet without ether header.
1813 	 */
1814 	m_adj(m, sizeof(struct ether_header));
1815 
1816 	/*
1817 	 * Find the packet's target msgport.
1818 	 */
1819 	port = ether_mport(isr, &m);
1820 	if (port == NULL) {
1821 		KKASSERT(m == NULL);
1822 		return;
1823 	}
1824 
1825 	/*
1826 	 * Restore ether header.
1827 	 */
1828 	if (save_eh != NULL) {
1829 		ether_restore_header(&m, eh, save_eh);
1830 		if (m == NULL)
1831 			return;
1832 	} else {
1833 		m->m_data -= ETHER_HDR_LEN;
1834 		m->m_len += ETHER_HDR_LEN;
1835 		m->m_pkthdr.len += ETHER_HDR_LEN;
1836 	}
1837 
1838 	/*
1839 	 * Initialize mbuf's netmsg packet _after_ possible
1840 	 * ether header restoration, else the initialized
1841 	 * netmsg packet may be lost during ether header
1842 	 * restoration.
1843 	 */
1844 	ether_init_netpacket(isr, m);
1845 
1846 #ifdef ETHER_INPUT_CHAIN
1847 	if (chain != NULL) {
1848 		struct mbuf_chain *c;
1849 		int cpuid;
1850 
1851 		m->m_pkthdr.header = port; /* XXX */
1852 		cpuid = port->mpu_td->td_gd->gd_cpuid;
1853 
1854 		c = &chain[cpuid];
1855 		if (c->mc_head == NULL) {
1856 			c->mc_head = c->mc_tail = m;
1857 		} else {
1858 			c->mc_tail->m_nextpkt = m;
1859 			c->mc_tail = m;
1860 		}
1861 		m->m_nextpkt = NULL;
1862 	} else
1863 #endif	/* ETHER_INPUT_CHAIN */
1864 		lwkt_sendmsg(port, &m->m_hdr.mh_netmsg.nm_netmsg.nm_lmsg);
1865 }
1866 
1867 #endif	/* ETHER_INPUT2 */
1868