xref: /dflybsd-src/sys/net/if_ethersubr.c (revision 5a975a3dcb66c383a644f58dd3a9a43c6ba43ac9)
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 1.74 2008/06/25 11:45:07 sephe 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_netgraph.h"
43 #include "opt_carp.h"
44 #include "opt_ethernet.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/globaldata.h>
49 #include <sys/kernel.h>
50 #include <sys/malloc.h>
51 #include <sys/mbuf.h>
52 #include <sys/msgport.h>
53 #include <sys/socket.h>
54 #include <sys/sockio.h>
55 #include <sys/sysctl.h>
56 #include <sys/thread.h>
57 #include <sys/thread2.h>
58 
59 #include <net/if.h>
60 #include <net/netisr.h>
61 #include <net/route.h>
62 #include <net/if_llc.h>
63 #include <net/if_dl.h>
64 #include <net/if_types.h>
65 #include <net/ifq_var.h>
66 #include <net/bpf.h>
67 #include <net/ethernet.h>
68 #include <net/vlan/if_vlan_ether.h>
69 #include <net/netmsg2.h>
70 
71 #if defined(INET) || defined(INET6)
72 #include <netinet/in.h>
73 #include <netinet/in_var.h>
74 #include <netinet/if_ether.h>
75 #include <net/ipfw/ip_fw.h>
76 #include <net/dummynet/ip_dummynet.h>
77 #endif
78 #ifdef INET6
79 #include <netinet6/nd6.h>
80 #endif
81 
82 #ifdef CARP
83 #include <netinet/ip_carp.h>
84 #endif
85 
86 #ifdef IPX
87 #include <netproto/ipx/ipx.h>
88 #include <netproto/ipx/ipx_if.h>
89 int (*ef_inputp)(struct ifnet*, const struct ether_header *eh, struct mbuf *m);
90 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst,
91 		  short *tp, int *hlen);
92 #endif
93 
94 #ifdef NS
95 #include <netns/ns.h>
96 #include <netns/ns_if.h>
97 ushort ns_nettype;
98 int ether_outputdebug = 0;
99 int ether_inputdebug = 0;
100 #endif
101 
102 #ifdef NETATALK
103 #include <netproto/atalk/at.h>
104 #include <netproto/atalk/at_var.h>
105 #include <netproto/atalk/at_extern.h>
106 
107 #define	llc_snap_org_code	llc_un.type_snap.org_code
108 #define	llc_snap_ether_type	llc_un.type_snap.ether_type
109 
110 extern u_char	at_org_code[3];
111 extern u_char	aarp_org_code[3];
112 #endif /* NETATALK */
113 
114 /* netgraph node hooks for ng_ether(4) */
115 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
116 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp,
117 		struct mbuf *m, const struct ether_header *eh);
118 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
119 void	(*ng_ether_attach_p)(struct ifnet *ifp);
120 void	(*ng_ether_detach_p)(struct ifnet *ifp);
121 
122 int	(*vlan_input_p)(struct mbuf *, struct mbuf_chain *);
123 void	(*vlan_input2_p)(struct mbuf *);
124 
125 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
126 			struct rtentry *);
127 static void ether_restore_header(struct mbuf **, const struct ether_header *,
128 				 const struct ether_header *);
129 static void ether_demux_chain(struct ifnet *, struct mbuf *,
130 			      struct mbuf_chain *);
131 
132 /*
133  * if_bridge support
134  */
135 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
136 int (*bridge_output_p)(struct ifnet *, struct mbuf *);
137 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
138 
139 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
140 			      struct sockaddr *);
141 
142 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
143 	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
144 };
145 
146 #define gotoerr(e) do { error = (e); goto bad; } while (0)
147 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
148 
149 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
150 				struct ip_fw **rule,
151 				const struct ether_header *eh);
152 
153 static int ether_ipfw;
154 static u_int ether_restore_hdr;
155 static u_int ether_prepend_hdr;
156 
157 SYSCTL_DECL(_net_link);
158 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
159 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
160 	   &ether_ipfw, 0, "Pass ether pkts through firewall");
161 SYSCTL_UINT(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW,
162 	    &ether_restore_hdr, 0, "# of ether header restoration");
163 SYSCTL_UINT(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW,
164 	    &ether_prepend_hdr, 0,
165 	    "# of ether header restoration which prepends mbuf");
166 
167 /*
168  * Ethernet output routine.
169  * Encapsulate a packet of type family for the local net.
170  * Use trailer local net encapsulation if enough data in first
171  * packet leaves a multiple of 512 bytes of data in remainder.
172  * Assumes that ifp is actually pointer to arpcom structure.
173  */
174 static int
175 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
176 	     struct rtentry *rt)
177 {
178 	struct ether_header *eh, *deh;
179 	u_char *edst;
180 	int loop_copy = 0;
181 	int hlen = ETHER_HDR_LEN;	/* link layer header length */
182 	struct arpcom *ac = IFP2AC(ifp);
183 	int error;
184 
185 	ASSERT_NOT_SERIALIZED(ifp->if_serializer);
186 
187 	if (ifp->if_flags & IFF_MONITOR)
188 		gotoerr(ENETDOWN);
189 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
190 		gotoerr(ENETDOWN);
191 
192 	M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT);
193 	if (m == NULL)
194 		return (ENOBUFS);
195 	eh = mtod(m, struct ether_header *);
196 	edst = eh->ether_dhost;
197 
198 	/*
199 	 * Fill in the destination ethernet address and frame type.
200 	 */
201 	switch (dst->sa_family) {
202 #ifdef INET
203 	case AF_INET:
204 		if (!arpresolve(ifp, rt, m, dst, edst))
205 			return (0);	/* if not yet resolved */
206 		eh->ether_type = htons(ETHERTYPE_IP);
207 		break;
208 #endif
209 #ifdef INET6
210 	case AF_INET6:
211 		if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst))
212 			return (0);		/* Something bad happenned. */
213 		eh->ether_type = htons(ETHERTYPE_IPV6);
214 		break;
215 #endif
216 #ifdef IPX
217 	case AF_IPX:
218 		if (ef_outputp != NULL) {
219 			error = ef_outputp(ifp, &m, dst, &eh->ether_type,
220 					   &hlen);
221 			if (error)
222 				goto bad;
223 		} else {
224 			eh->ether_type = htons(ETHERTYPE_IPX);
225 			bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
226 			      edst, ETHER_ADDR_LEN);
227 		}
228 		break;
229 #endif
230 #ifdef NETATALK
231 	case AF_APPLETALK: {
232 		struct at_ifaddr *aa;
233 
234 		if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL) {
235 			error = 0;	/* XXX */
236 			goto bad;
237 		}
238 		/*
239 		 * In the phase 2 case, need to prepend an mbuf for
240 		 * the llc header.  Since we must preserve the value
241 		 * of m, which is passed to us by value, we m_copy()
242 		 * the first mbuf, and use it for our llc header.
243 		 */
244 		if (aa->aa_flags & AFA_PHASE2) {
245 			struct llc llc;
246 
247 			M_PREPEND(m, sizeof(struct llc), MB_DONTWAIT);
248 			eh = mtod(m, struct ether_header *);
249 			edst = eh->ether_dhost;
250 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
251 			llc.llc_control = LLC_UI;
252 			bcopy(at_org_code, llc.llc_snap_org_code,
253 			      sizeof at_org_code);
254 			llc.llc_snap_ether_type = htons(ETHERTYPE_AT);
255 			bcopy(&llc,
256 			      mtod(m, caddr_t) + sizeof(struct ether_header),
257 			      sizeof(struct llc));
258 			eh->ether_type = htons(m->m_pkthdr.len);
259 			hlen = sizeof(struct llc) + ETHER_HDR_LEN;
260 		} else {
261 			eh->ether_type = htons(ETHERTYPE_AT);
262 		}
263 		if (!aarpresolve(ac, m, (struct sockaddr_at *)dst, edst))
264 			return (0);
265 		break;
266 	  }
267 #endif
268 #ifdef NS
269 	case AF_NS:
270 		switch(ns_nettype) {
271 		default:
272 		case 0x8137:	/* Novell Ethernet_II Ethernet TYPE II */
273 			eh->ether_type = 0x8137;
274 			break;
275 		case 0x0:	/* Novell 802.3 */
276 			eh->ether_type = htons(m->m_pkthdr.len);
277 			break;
278 		case 0xe0e0:	/* Novell 802.2 and Token-Ring */
279 			M_PREPEND(m, 3, MB_DONTWAIT);
280 			eh = mtod(m, struct ether_header *);
281 			edst = eh->ether_dhost;
282 			eh->ether_type = htons(m->m_pkthdr.len);
283 			cp = mtod(m, u_char *) + sizeof(struct ether_header);
284 			*cp++ = 0xE0;
285 			*cp++ = 0xE0;
286 			*cp++ = 0x03;
287 			break;
288 		}
289 		bcopy(&(((struct sockaddr_ns *)dst)->sns_addr.x_host), edst,
290 		      ETHER_ADDR_LEN);
291 		/*
292 		 * XXX if ns_thishost is the same as the node's ethernet
293 		 * address then just the default code will catch this anyhow.
294 		 * So I'm not sure if this next clause should be here at all?
295 		 * [JRE]
296 		 */
297 		if (bcmp(edst, &ns_thishost, ETHER_ADDR_LEN) == 0) {
298 			m->m_pkthdr.rcvif = ifp;
299 			netisr_dispatch(NETISR_NS, m);
300 			return (error);
301 		}
302 		if (bcmp(edst, &ns_broadhost, ETHER_ADDR_LEN) == 0)
303 			m->m_flags |= M_BCAST;
304 		break;
305 #endif
306 	case pseudo_AF_HDRCMPLT:
307 	case AF_UNSPEC:
308 		loop_copy = -1; /* if this is for us, don't do it */
309 		deh = (struct ether_header *)dst->sa_data;
310 		memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
311 		eh->ether_type = deh->ether_type;
312 		break;
313 
314 	default:
315 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
316 		gotoerr(EAFNOSUPPORT);
317 	}
318 
319 	if (dst->sa_family == pseudo_AF_HDRCMPLT)	/* unlikely */
320 		memcpy(eh->ether_shost,
321 		       ((struct ether_header *)dst->sa_data)->ether_shost,
322 		       ETHER_ADDR_LEN);
323 	else
324 		memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
325 
326 	/*
327 	 * Bridges require special output handling.
328 	 */
329 	if (ifp->if_bridge) {
330 		KASSERT(bridge_output_p != NULL,
331 			("%s: if_bridge not loaded!", __func__));
332 		return bridge_output_p(ifp, m);
333 	}
334 
335 	/*
336 	 * If a simplex interface, and the packet is being sent to our
337 	 * Ethernet address or a broadcast address, loopback a copy.
338 	 * XXX To make a simplex device behave exactly like a duplex
339 	 * device, we should copy in the case of sending to our own
340 	 * ethernet address (thus letting the original actually appear
341 	 * on the wire). However, we don't do that here for security
342 	 * reasons and compatibility with the original behavior.
343 	 */
344 	if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
345 		int csum_flags = 0;
346 
347 		if (m->m_pkthdr.csum_flags & CSUM_IP)
348 			csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
349 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
350 			csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
351 		if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
352 			struct mbuf *n;
353 
354 			if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) {
355 				n->m_pkthdr.csum_flags |= csum_flags;
356 				if (csum_flags & CSUM_DATA_VALID)
357 					n->m_pkthdr.csum_data = 0xffff;
358 				if_simloop(ifp, n, dst->sa_family, hlen);
359 			} else
360 				ifp->if_iqdrops++;
361 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
362 				ETHER_ADDR_LEN) == 0) {
363 			m->m_pkthdr.csum_flags |= csum_flags;
364 			if (csum_flags & CSUM_DATA_VALID)
365 				m->m_pkthdr.csum_data = 0xffff;
366 			if_simloop(ifp, m, dst->sa_family, hlen);
367 			return (0);	/* XXX */
368 		}
369 	}
370 
371 #ifdef CARP
372 	if (ifp->if_carp && (error = carp_output(ifp, m, dst, NULL)))
373 		goto bad;
374 #endif
375 
376 
377 	/* Handle ng_ether(4) processing, if any */
378 	if (ng_ether_output_p != NULL) {
379 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0)
380 			goto bad;
381 		if (m == NULL)
382 			return (0);
383 	}
384 
385 	/* Continue with link-layer output */
386 	return ether_output_frame(ifp, m);
387 
388 bad:
389 	m_freem(m);
390 	return (error);
391 }
392 
393 /*
394  * Ethernet link layer output routine to send a raw frame to the device.
395  *
396  * This assumes that the 14 byte Ethernet header is present and contiguous
397  * in the first mbuf.
398  */
399 int
400 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
401 {
402 	struct ip_fw *rule = NULL;
403 	int error = 0;
404 	struct altq_pktattr pktattr;
405 	struct m_tag *mtag;
406 
407 	ASSERT_NOT_SERIALIZED(ifp->if_serializer);
408 
409 	/* Extract info from dummynet tag */
410 	mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
411 	if (mtag != NULL) {
412 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
413 
414 		m_tag_delete(m, mtag);
415 		mtag = NULL;
416 	}
417 
418 	if (ifq_is_enabled(&ifp->if_snd))
419 		altq_etherclassify(&ifp->if_snd, m, &pktattr);
420 	crit_enter();
421 	if (IPFW_LOADED && ether_ipfw != 0) {
422 		struct ether_header save_eh, *eh;
423 
424 		eh = mtod(m, struct ether_header *);
425 		save_eh = *eh;
426 		m_adj(m, ETHER_HDR_LEN);
427 		if (!ether_ipfw_chk(&m, ifp, &rule, eh)) {
428 			crit_exit();
429 			if (m != NULL) {
430 				m_freem(m);
431 				return ENOBUFS; /* pkt dropped */
432 			} else
433 				return 0;	/* consumed e.g. in a pipe */
434 		}
435 
436 		/* packet was ok, restore the ethernet header */
437 		ether_restore_header(&m, eh, &save_eh);
438 		if (m == NULL) {
439 			crit_exit();
440 			return ENOBUFS;
441 		}
442 	}
443 	crit_exit();
444 
445 	/*
446 	 * Queue message on interface, update output statistics if
447 	 * successful, and start output if interface not yet active.
448 	 */
449 	error = ifq_dispatch(ifp, m, &pktattr);
450 	return (error);
451 }
452 
453 /*
454  * ipfw processing for ethernet packets (in and out).
455  * The second parameter is NULL from ether_demux(), and ifp from
456  * ether_output_frame().
457  */
458 static boolean_t
459 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule,
460 	       const struct ether_header *eh)
461 {
462 	struct ether_header save_eh = *eh;	/* might be a ptr in m */
463 	struct ip_fw_args args;
464 	struct m_tag *mtag;
465 	int i;
466 
467 	if (*rule != NULL && fw_one_pass)
468 		return TRUE; /* dummynet packet, already partially processed */
469 
470 	/*
471 	 * I need some amount of data to be contiguous.
472 	 */
473 	i = min((*m0)->m_pkthdr.len, max_protohdr);
474 	if ((*m0)->m_len < i) {
475 		*m0 = m_pullup(*m0, i);
476 		if (*m0 == NULL)
477 			return FALSE;
478 	}
479 
480 	args.m = *m0;		/* the packet we are looking at		*/
481 	args.oif = dst;		/* destination, if any			*/
482 	if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
483 		m_tag_delete(*m0, mtag);
484 	args.rule = *rule;	/* matching rule to restart		*/
485 	args.next_hop = NULL;	/* we do not support forward yet	*/
486 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
487 	i = ip_fw_chk_ptr(&args);
488 	*m0 = args.m;
489 	*rule = args.rule;
490 
491 	if ((i & IP_FW_PORT_DENY_FLAG) || *m0 == NULL)	/* drop */
492 		return FALSE;
493 
494 	if (i == 0)					/* a PASS rule.  */
495 		return TRUE;
496 
497 	if (i & IP_FW_PORT_DYNT_FLAG) {
498 		/*
499 		 * Pass the pkt to dummynet, which consumes it.
500 		 */
501 		struct mbuf *m;
502 
503 		m = *m0;	/* pass the original to dummynet */
504 		*m0 = NULL;	/* and nothing back to the caller */
505 
506 		ether_restore_header(&m, eh, &save_eh);
507 		if (m == NULL)
508 			return FALSE;
509 
510 		ip_fw_dn_io_ptr(m, (i & 0xffff),
511 			dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
512 		return FALSE;
513 	}
514 	/*
515 	 * XXX at some point add support for divert/forward actions.
516 	 * If none of the above matches, we have to drop the pkt.
517 	 */
518 	return FALSE;
519 }
520 
521 /*
522  * Process a received Ethernet packet.
523  *
524  * The ethernet header is assumed to be in the mbuf so the caller
525  * MUST MAKE SURE that there are at least sizeof(struct ether_header)
526  * bytes in the first mbuf.
527  *
528  * This allows us to concentrate in one place a bunch of code which
529  * is replicated in all device drivers. Also, many functions called
530  * from ether_input() try to put the eh back into the mbuf, so we
531  * can later propagate the 'contiguous packet' interface to them.
532  *
533  * NOTA BENE: for all drivers "eh" is a pointer into the first mbuf or
534  * cluster, right before m_data. So be very careful when working on m,
535  * as you could destroy *eh !!
536  *
537  * First we perform any link layer operations, then continue to the
538  * upper layers with ether_demux().
539  */
540 void
541 ether_input_chain(struct ifnet *ifp, struct mbuf *m, struct mbuf_chain *chain)
542 {
543 	struct ether_header *eh;
544 
545 	ASSERT_SERIALIZED(ifp->if_serializer);
546 	M_ASSERTPKTHDR(m);
547 
548 	/* Discard packet if interface is not up */
549 	if (!(ifp->if_flags & IFF_UP)) {
550 		m_freem(m);
551 		return;
552 	}
553 
554 	if (m->m_len < sizeof(struct ether_header)) {
555 		/* XXX error in the caller. */
556 		m_freem(m);
557 		return;
558 	}
559 	eh = mtod(m, struct ether_header *);
560 
561 	if (ntohs(eh->ether_type) == ETHERTYPE_VLAN &&
562 	    (m->m_flags & M_VLANTAG) == 0) {
563 		/*
564 		 * Extract vlan tag if hardware does not do it for us
565 		 */
566 		vlan_ether_decap(&m);
567 		if (m == NULL)
568 			return;
569 		eh = mtod(m, struct ether_header *);
570 	}
571 
572 	m->m_pkthdr.rcvif = ifp;
573 
574 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
575 		if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
576 			 ifp->if_addrlen) == 0)
577 			m->m_flags |= M_BCAST;
578 		else
579 			m->m_flags |= M_MCAST;
580 		ifp->if_imcasts++;
581 	}
582 
583 	ETHER_BPF_MTAP(ifp, m);
584 
585 	ifp->if_ibytes += m->m_pkthdr.len;
586 
587 	if (ifp->if_flags & IFF_MONITOR) {
588 		/*
589 		 * Interface marked for monitoring; discard packet.
590 		 */
591 		 m_freem(m);
592 		 return;
593 	}
594 
595 	/*
596 	 * Tap the packet off here for a bridge.  bridge_input()
597 	 * will return NULL if it has consumed the packet, otherwise
598 	 * it gets processed as normal.  Note that bridge_input()
599 	 * will always return the original packet if we need to
600 	 * process it locally.
601 	 */
602 	if (ifp->if_bridge) {
603 		KASSERT(bridge_input_p != NULL,
604 			("%s: if_bridge not loaded!", __func__));
605 
606 		if(m->m_flags & M_PROTO1) {
607 			m->m_flags &= ~M_PROTO1;
608 		} else {
609 			/* clear M_PROMISC, in case the packets comes from a vlan */
610 			/* m->m_flags &= ~M_PROMISC; */
611 			lwkt_serialize_exit(ifp->if_serializer);
612 			m = bridge_input_p(ifp, m);
613 			lwkt_serialize_enter(ifp->if_serializer);
614 			if (m == NULL)
615 				return;
616 
617 			KASSERT(ifp == m->m_pkthdr.rcvif,
618 				("bridge_input_p changed rcvif\n"));
619 
620 			/* 'm' may be changed by bridge_input_p() */
621 			eh = mtod(m, struct ether_header *);
622 		}
623 	}
624 
625 	/* Handle ng_ether(4) processing, if any */
626 	if (ng_ether_input_p != NULL) {
627 		ng_ether_input_p(ifp, &m);
628 		if (m == NULL)
629 			return;
630 
631 		/* 'm' may be changed by ng_ether_input_p() */
632 		eh = mtod(m, struct ether_header *);
633 	}
634 
635 	/* Continue with upper layer processing */
636 	ether_demux_chain(ifp, m, chain);
637 }
638 
639 void
640 ether_input(struct ifnet *ifp, struct mbuf *m)
641 {
642 	ether_input_chain(ifp, m, NULL);
643 }
644 
645 /*
646  * Upper layer processing for a received Ethernet packet.
647  */
648 static void
649 ether_demux_chain(struct ifnet *ifp, struct mbuf *m, struct mbuf_chain *chain)
650 {
651 	struct ether_header save_eh, *eh;
652 	int isr;
653 	u_short ether_type;
654 	struct ip_fw *rule = NULL;
655 	struct m_tag *mtag;
656 #ifdef NETATALK
657 	struct llc *l;
658 #endif
659 
660 	M_ASSERTPKTHDR(m);
661 	KASSERT(m->m_len >= ETHER_HDR_LEN,
662 		("ether header is no contiguous!\n"));
663 
664 	eh = mtod(m, struct ether_header *);
665 	save_eh = *eh;
666 
667 	/* XXX old crufty stuff, needs to be removed */
668 	m_adj(m, sizeof(struct ether_header));
669 
670 	/* Extract info from dummynet tag */
671 	mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
672 	if (mtag != NULL) {
673 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
674 		KKASSERT(ifp == NULL);
675 		ifp = m->m_pkthdr.rcvif;
676 
677 		m_tag_delete(m, mtag);
678 		mtag = NULL;
679 	}
680 	if (rule)	/* packet is passing the second time */
681 		goto post_stats;
682 
683 #ifdef CARP
684 	/*
685 	 * XXX: Okay, we need to call carp_forus() and - if it is for
686 	 * us jump over code that does the normal check
687 	 * "ac_enaddr == ether_dhost". The check sequence is a bit
688 	 * different from OpenBSD, so we jump over as few code as
689 	 * possible, to catch _all_ sanity checks. This needs
690 	 * evaluation, to see if the carp ether_dhost values break any
691 	 * of these checks!
692 	 */
693 	if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost))
694 		goto post_stats;
695 #endif
696 
697 	/*
698 	 * Discard packet if upper layers shouldn't see it because
699 	 * it was unicast to a different Ethernet address.  If the
700 	 * driver is working properly, then this situation can only
701 	 * happen when the interface is in promiscuous mode.
702 	 */
703 	if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
704 	    (eh->ether_dhost[0] & 1) == 0 &&
705 	    bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
706 		m_freem(m);
707 		return;
708 	}
709 
710 post_stats:
711 	if (IPFW_LOADED && ether_ipfw != 0) {
712 		if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
713 			m_freem(m);
714 			return;
715 		}
716 	}
717 
718 	ether_type = ntohs(save_eh.ether_type);
719 
720 	if (m->m_flags & M_VLANTAG) {
721 		if (ether_type == ETHERTYPE_VLAN) {
722 			/*
723 			 * To prevent possible dangerous recursion,
724 			 * we don't do vlan-in-vlan
725 			 */
726 			m->m_pkthdr.rcvif->if_noproto++;
727 			m_freem(m);
728 			return;
729 		}
730 
731 		if (vlan_input_p != NULL) {
732 			ether_restore_header(&m, eh, &save_eh);
733 			if (m != NULL)
734 				vlan_input_p(m, chain);
735 		} else {
736 			m->m_pkthdr.rcvif->if_noproto++;
737 			m_freem(m);
738 		}
739 		return;
740 	}
741 	KKASSERT(ether_type != ETHERTYPE_VLAN);
742 
743 	switch (ether_type) {
744 #ifdef INET
745 	case ETHERTYPE_IP:
746 		if (ipflow_fastforward(m, ifp->if_serializer))
747 			return;
748 		isr = NETISR_IP;
749 		break;
750 
751 	case ETHERTYPE_ARP:
752 		if (ifp->if_flags & IFF_NOARP) {
753 			/* Discard packet if ARP is disabled on interface */
754 			m_freem(m);
755 			return;
756 		}
757 		isr = NETISR_ARP;
758 		break;
759 #endif
760 
761 #ifdef INET6
762 	case ETHERTYPE_IPV6:
763 		isr = NETISR_IPV6;
764 		break;
765 #endif
766 
767 #ifdef IPX
768 	case ETHERTYPE_IPX:
769 		if (ef_inputp && ef_inputp(ifp, &save_eh, m) == 0)
770 			return;
771 		isr = NETISR_IPX;
772 		break;
773 #endif
774 
775 #ifdef NS
776 	case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
777 		isr = NETISR_NS;
778 		break;
779 
780 #endif
781 
782 #ifdef NETATALK
783 	case ETHERTYPE_AT:
784 		isr = NETISR_ATALK1;
785 		break;
786 	case ETHERTYPE_AARP:
787 		isr = NETISR_AARP;
788 		break;
789 #endif
790 
791 	default:
792 #ifdef IPX
793 		if (ef_inputp && ef_inputp(ifp, &save_eh, m) == 0)
794 			return;
795 #endif
796 #ifdef NS
797 		checksum = mtod(m, ushort *);
798 		/* Novell 802.3 */
799 		if ((ether_type <= ETHERMTU) &&
800 		    ((*checksum == 0xffff) || (*checksum == 0xE0E0))) {
801 			if (*checksum == 0xE0E0) {
802 				m->m_pkthdr.len -= 3;
803 				m->m_len -= 3;
804 				m->m_data += 3;
805 			}
806 			isr = NETISR_NS;
807 			break;
808 		}
809 #endif
810 #ifdef NETATALK
811 		if (ether_type > ETHERMTU)
812 			goto dropanyway;
813 		l = mtod(m, struct llc *);
814 		if (l->llc_dsap == LLC_SNAP_LSAP &&
815 		    l->llc_ssap == LLC_SNAP_LSAP &&
816 		    l->llc_control == LLC_UI) {
817 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
818 				 sizeof at_org_code) == 0 &&
819 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
820 				m_adj(m, sizeof(struct llc));
821 				isr = NETISR_ATALK2;
822 				break;
823 			}
824 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
825 				 sizeof aarp_org_code) == 0 &&
826 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
827 				m_adj(m, sizeof(struct llc));
828 				isr = NETISR_AARP;
829 				break;
830 			}
831 		}
832 dropanyway:
833 #endif
834 		if (ng_ether_input_orphan_p != NULL)
835 			(*ng_ether_input_orphan_p)(ifp, m, &save_eh);
836 		else
837 			m_freem(m);
838 		return;
839 	}
840 
841 #ifdef ETHER_INPUT_CHAIN
842 	if (chain != NULL) {
843 		struct mbuf_chain *c;
844 		lwkt_port_t port;
845 		int cpuid;
846 
847 		port = netisr_mport(isr, &m);
848 		if (port == NULL)
849 			return;
850 
851 		m->m_pkthdr.header = port; /* XXX */
852 		cpuid = port->mpu_td->td_gd->gd_cpuid;
853 
854 		c = &chain[cpuid];
855 		if (c->mc_head == NULL) {
856 			c->mc_head = c->mc_tail = m;
857 		} else {
858 			c->mc_tail->m_nextpkt = m;
859 			c->mc_tail = m;
860 		}
861 		m->m_nextpkt = NULL;
862 	} else
863 #endif	/* ETHER_INPUT_CHAIN */
864 		netisr_dispatch(isr, m);
865 }
866 
867 void
868 ether_demux(struct ifnet *ifp, struct mbuf *m)
869 {
870 	ether_demux_chain(ifp, m, NULL);
871 }
872 
873 /*
874  * Perform common duties while attaching to interface list
875  */
876 
877 void
878 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer)
879 {
880 	ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
881 			   serializer);
882 }
883 
884 void
885 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen,
886 		   lwkt_serialize_t serializer)
887 {
888 	struct sockaddr_dl *sdl;
889 
890 	ifp->if_type = IFT_ETHER;
891 	ifp->if_addrlen = ETHER_ADDR_LEN;
892 	ifp->if_hdrlen = ETHER_HDR_LEN;
893 	if_attach(ifp, serializer);
894 	ifp->if_mtu = ETHERMTU;
895 	if (ifp->if_baudrate == 0)
896 		ifp->if_baudrate = 10000000;
897 	ifp->if_output = ether_output;
898 	ifp->if_input = ether_input;
899 	ifp->if_resolvemulti = ether_resolvemulti;
900 	ifp->if_broadcastaddr = etherbroadcastaddr;
901 	sdl = IF_LLSOCKADDR(ifp);
902 	sdl->sdl_type = IFT_ETHER;
903 	sdl->sdl_alen = ifp->if_addrlen;
904 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
905 	/*
906 	 * XXX Keep the current drivers happy.
907 	 * XXX Remove once all drivers have been cleaned up
908 	 */
909 	if (lla != IFP2AC(ifp)->ac_enaddr)
910 		bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
911 	bpfattach(ifp, dlt, hdrlen);
912 	if (ng_ether_attach_p != NULL)
913 		(*ng_ether_attach_p)(ifp);
914 
915 	if_printf(ifp, "MAC address: %6D\n", lla, ":");
916 }
917 
918 /*
919  * Perform common duties while detaching an Ethernet interface
920  */
921 void
922 ether_ifdetach(struct ifnet *ifp)
923 {
924 	if_down(ifp);
925 
926 	if (ng_ether_detach_p != NULL)
927 		(*ng_ether_detach_p)(ifp);
928 	bpfdetach(ifp);
929 	if_detach(ifp);
930 }
931 
932 int
933 ether_ioctl(struct ifnet *ifp, int command, caddr_t data)
934 {
935 	struct ifaddr *ifa = (struct ifaddr *) data;
936 	struct ifreq *ifr = (struct ifreq *) data;
937 	int error = 0;
938 
939 #define IF_INIT(ifp) \
940 do { \
941 	if (((ifp)->if_flags & IFF_UP) == 0) { \
942 		(ifp)->if_flags |= IFF_UP; \
943 		(ifp)->if_init((ifp)->if_softc); \
944 	} \
945 } while (0)
946 
947 	ASSERT_SERIALIZED(ifp->if_serializer);
948 
949 	switch (command) {
950 	case SIOCSIFADDR:
951 		switch (ifa->ifa_addr->sa_family) {
952 #ifdef INET
953 		case AF_INET:
954 			IF_INIT(ifp);	/* before arpwhohas */
955 			arp_ifinit(ifp, ifa);
956 			break;
957 #endif
958 #ifdef IPX
959 		/*
960 		 * XXX - This code is probably wrong
961 		 */
962 		case AF_IPX:
963 			{
964 			struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr;
965 			struct arpcom *ac = IFP2AC(ifp);
966 
967 			if (ipx_nullhost(*ina))
968 				ina->x_host = *(union ipx_host *) ac->ac_enaddr;
969 			else
970 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
971 				      sizeof ac->ac_enaddr);
972 
973 			IF_INIT(ifp);	/* Set new address. */
974 			break;
975 			}
976 #endif
977 #ifdef NS
978 		/*
979 		 * XXX - This code is probably wrong
980 		 */
981 		case AF_NS:
982 		{
983 			struct ns_addr *ina = &(IA_SNS(ifa)->sns_addr);
984 			struct arpcom *ac = IFP2AC(ifp);
985 
986 			if (ns_nullhost(*ina))
987 				ina->x_host = *(union ns_host *)(ac->ac_enaddr);
988 			else
989 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
990 				      sizeof ac->ac_enaddr);
991 
992 			/*
993 			 * Set new address
994 			 */
995 			IF_INIT(ifp);
996 			break;
997 		}
998 #endif
999 		default:
1000 			IF_INIT(ifp);
1001 			break;
1002 		}
1003 		break;
1004 
1005 	case SIOCGIFADDR:
1006 		bcopy(IFP2AC(ifp)->ac_enaddr,
1007 		      ((struct sockaddr *)ifr->ifr_data)->sa_data,
1008 		      ETHER_ADDR_LEN);
1009 		break;
1010 
1011 	case SIOCSIFMTU:
1012 		/*
1013 		 * Set the interface MTU.
1014 		 */
1015 		if (ifr->ifr_mtu > ETHERMTU) {
1016 			error = EINVAL;
1017 		} else {
1018 			ifp->if_mtu = ifr->ifr_mtu;
1019 		}
1020 		break;
1021 	default:
1022 		error = EINVAL;
1023 		break;
1024 	}
1025 	return (error);
1026 
1027 #undef IF_INIT
1028 }
1029 
1030 int
1031 ether_resolvemulti(
1032 	struct ifnet *ifp,
1033 	struct sockaddr **llsa,
1034 	struct sockaddr *sa)
1035 {
1036 	struct sockaddr_dl *sdl;
1037 	struct sockaddr_in *sin;
1038 #ifdef INET6
1039 	struct sockaddr_in6 *sin6;
1040 #endif
1041 	u_char *e_addr;
1042 
1043 	switch(sa->sa_family) {
1044 	case AF_LINK:
1045 		/*
1046 		 * No mapping needed. Just check that it's a valid MC address.
1047 		 */
1048 		sdl = (struct sockaddr_dl *)sa;
1049 		e_addr = LLADDR(sdl);
1050 		if ((e_addr[0] & 1) != 1)
1051 			return EADDRNOTAVAIL;
1052 		*llsa = 0;
1053 		return 0;
1054 
1055 #ifdef INET
1056 	case AF_INET:
1057 		sin = (struct sockaddr_in *)sa;
1058 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1059 			return EADDRNOTAVAIL;
1060 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
1061 		       M_WAITOK | M_ZERO);
1062 		sdl->sdl_len = sizeof *sdl;
1063 		sdl->sdl_family = AF_LINK;
1064 		sdl->sdl_index = ifp->if_index;
1065 		sdl->sdl_type = IFT_ETHER;
1066 		sdl->sdl_alen = ETHER_ADDR_LEN;
1067 		e_addr = LLADDR(sdl);
1068 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1069 		*llsa = (struct sockaddr *)sdl;
1070 		return 0;
1071 #endif
1072 #ifdef INET6
1073 	case AF_INET6:
1074 		sin6 = (struct sockaddr_in6 *)sa;
1075 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1076 			/*
1077 			 * An IP6 address of 0 means listen to all
1078 			 * of the Ethernet multicast address used for IP6.
1079 			 * (This is used for multicast routers.)
1080 			 */
1081 			ifp->if_flags |= IFF_ALLMULTI;
1082 			*llsa = 0;
1083 			return 0;
1084 		}
1085 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1086 			return EADDRNOTAVAIL;
1087 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
1088 		       M_WAITOK | M_ZERO);
1089 		sdl->sdl_len = sizeof *sdl;
1090 		sdl->sdl_family = AF_LINK;
1091 		sdl->sdl_index = ifp->if_index;
1092 		sdl->sdl_type = IFT_ETHER;
1093 		sdl->sdl_alen = ETHER_ADDR_LEN;
1094 		e_addr = LLADDR(sdl);
1095 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1096 		*llsa = (struct sockaddr *)sdl;
1097 		return 0;
1098 #endif
1099 
1100 	default:
1101 		/*
1102 		 * Well, the text isn't quite right, but it's the name
1103 		 * that counts...
1104 		 */
1105 		return EAFNOSUPPORT;
1106 	}
1107 }
1108 
1109 #if 0
1110 /*
1111  * This is for reference.  We have a table-driven version
1112  * of the little-endian crc32 generator, which is faster
1113  * than the double-loop.
1114  */
1115 uint32_t
1116 ether_crc32_le(const uint8_t *buf, size_t len)
1117 {
1118 	uint32_t c, crc, carry;
1119 	size_t i, j;
1120 
1121 	crc = 0xffffffffU;	/* initial value */
1122 
1123 	for (i = 0; i < len; i++) {
1124 		c = buf[i];
1125 		for (j = 0; j < 8; j++) {
1126 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1127 			crc >>= 1;
1128 			c >>= 1;
1129 			if (carry)
1130 				crc = (crc ^ ETHER_CRC_POLY_LE);
1131 		}
1132 	}
1133 
1134 	return (crc);
1135 }
1136 #else
1137 uint32_t
1138 ether_crc32_le(const uint8_t *buf, size_t len)
1139 {
1140 	static const uint32_t crctab[] = {
1141 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1142 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1143 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1144 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1145 	};
1146 	uint32_t crc;
1147 	size_t i;
1148 
1149 	crc = 0xffffffffU;	/* initial value */
1150 
1151 	for (i = 0; i < len; i++) {
1152 		crc ^= buf[i];
1153 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1154 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1155 	}
1156 
1157 	return (crc);
1158 }
1159 #endif
1160 
1161 uint32_t
1162 ether_crc32_be(const uint8_t *buf, size_t len)
1163 {
1164 	uint32_t c, crc, carry;
1165 	size_t i, j;
1166 
1167 	crc = 0xffffffffU;	/* initial value */
1168 
1169 	for (i = 0; i < len; i++) {
1170 		c = buf[i];
1171 		for (j = 0; j < 8; j++) {
1172 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1173 			crc <<= 1;
1174 			c >>= 1;
1175 			if (carry)
1176 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1177 		}
1178 	}
1179 
1180 	return (crc);
1181 }
1182 
1183 /*
1184  * find the size of ethernet header, and call classifier
1185  */
1186 void
1187 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
1188 		   struct altq_pktattr *pktattr)
1189 {
1190 	struct ether_header *eh;
1191 	uint16_t ether_type;
1192 	int hlen, af, hdrsize;
1193 	caddr_t hdr;
1194 
1195 	hlen = sizeof(struct ether_header);
1196 	eh = mtod(m, struct ether_header *);
1197 
1198 	ether_type = ntohs(eh->ether_type);
1199 	if (ether_type < ETHERMTU) {
1200 		/* ick! LLC/SNAP */
1201 		struct llc *llc = (struct llc *)(eh + 1);
1202 		hlen += 8;
1203 
1204 		if (m->m_len < hlen ||
1205 		    llc->llc_dsap != LLC_SNAP_LSAP ||
1206 		    llc->llc_ssap != LLC_SNAP_LSAP ||
1207 		    llc->llc_control != LLC_UI)
1208 			goto bad;  /* not snap! */
1209 
1210 		ether_type = ntohs(llc->llc_un.type_snap.ether_type);
1211 	}
1212 
1213 	if (ether_type == ETHERTYPE_IP) {
1214 		af = AF_INET;
1215 		hdrsize = 20;  /* sizeof(struct ip) */
1216 #ifdef INET6
1217 	} else if (ether_type == ETHERTYPE_IPV6) {
1218 		af = AF_INET6;
1219 		hdrsize = 40;  /* sizeof(struct ip6_hdr) */
1220 #endif
1221 	} else
1222 		goto bad;
1223 
1224 	while (m->m_len <= hlen) {
1225 		hlen -= m->m_len;
1226 		m = m->m_next;
1227 	}
1228 	hdr = m->m_data + hlen;
1229 	if (m->m_len < hlen + hdrsize) {
1230 		/*
1231 		 * ip header is not in a single mbuf.  this should not
1232 		 * happen in the current code.
1233 		 * (todo: use m_pulldown in the future)
1234 		 */
1235 		goto bad;
1236 	}
1237 	m->m_data += hlen;
1238 	m->m_len -= hlen;
1239 	ifq_classify(ifq, m, af, pktattr);
1240 	m->m_data -= hlen;
1241 	m->m_len += hlen;
1242 
1243 	return;
1244 
1245 bad:
1246 	pktattr->pattr_class = NULL;
1247 	pktattr->pattr_hdr = NULL;
1248 	pktattr->pattr_af = AF_UNSPEC;
1249 }
1250 
1251 static void
1252 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
1253 		     const struct ether_header *save_eh)
1254 {
1255 	struct mbuf *m = *m0;
1256 
1257 	ether_restore_hdr++;
1258 
1259 	/*
1260 	 * Prepend the header, optimize for the common case of
1261 	 * eh pointing into the mbuf.
1262 	 */
1263 	if ((const void *)(eh + 1) == (void *)m->m_data) {
1264 		m->m_data -= ETHER_HDR_LEN;
1265 		m->m_len += ETHER_HDR_LEN;
1266 		m->m_pkthdr.len += ETHER_HDR_LEN;
1267 	} else {
1268 		ether_prepend_hdr++;
1269 
1270 		M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
1271 		if (m != NULL) {
1272 			bcopy(save_eh, mtod(m, struct ether_header *),
1273 			      ETHER_HDR_LEN);
1274 		}
1275 	}
1276 	*m0 = m;
1277 }
1278 
1279 #ifdef ETHER_INPUT_CHAIN
1280 
1281 static void
1282 ether_input_ipifunc(void *arg)
1283 {
1284 	struct mbuf *m, *next;
1285 	lwkt_port_t port;
1286 
1287 	m = arg;
1288 	do {
1289 		next = m->m_nextpkt;
1290 		m->m_nextpkt = NULL;
1291 
1292 		port = m->m_pkthdr.header;
1293 		m->m_pkthdr.header = NULL;
1294 
1295 		lwkt_sendmsg(port,
1296 		&m->m_hdr.mh_netmsg.nm_netmsg.nm_lmsg);
1297 
1298 		m = next;
1299 	} while (m != NULL);
1300 }
1301 
1302 void
1303 ether_input_dispatch(struct mbuf_chain *chain)
1304 {
1305 #ifdef SMP
1306 	int i;
1307 
1308 	for (i = 0; i < ncpus; ++i) {
1309 		if (chain[i].mc_head != NULL) {
1310 			lwkt_send_ipiq(globaldata_find(i),
1311 			ether_input_ipifunc, chain[i].mc_head);
1312 		}
1313 	}
1314 #else
1315 	if (chain->mc_head != NULL)
1316 		ether_input_ipifunc(chain->mc_head);
1317 #endif
1318 }
1319 
1320 void
1321 ether_input_chain_init(struct mbuf_chain *chain)
1322 {
1323 #ifdef SMP
1324 	int i;
1325 
1326 	for (i = 0; i < ncpus; ++i)
1327 		chain[i].mc_head = chain[i].mc_tail = NULL;
1328 #else
1329 	chain->mc_head = chain->mc_tail = NULL;
1330 #endif
1331 }
1332 
1333 #endif	/* ETHER_INPUT_CHAIN */
1334 
1335 #ifdef ETHER_INPUT2
1336 
1337 static void
1338 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
1339 {
1340 	struct ether_header *eh;
1341 	int isr, redispatch;
1342 	u_short ether_type;
1343 	struct ip_fw *rule = NULL;
1344 	struct m_tag *mtag;
1345 #ifdef NETATALK
1346 	struct llc *l;
1347 #endif
1348 
1349 	M_ASSERTPKTHDR(m);
1350 	KASSERT(m->m_len >= ETHER_HDR_LEN,
1351 		("ether header is no contiguous!\n"));
1352 
1353 	eh = mtod(m, struct ether_header *);
1354 
1355 	/* Extract info from dummynet tag */
1356 	mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
1357 	if (mtag != NULL) {
1358 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
1359 		KKASSERT(ifp == NULL);
1360 		ifp = m->m_pkthdr.rcvif;
1361 
1362 		m_tag_delete(m, mtag);
1363 		mtag = NULL;
1364 	}
1365 	if (rule)	/* packet is passing the second time */
1366 		goto post_stats;
1367 
1368 #ifdef CARP
1369 	/*
1370 	 * XXX: Okay, we need to call carp_forus() and - if it is for
1371 	 * us jump over code that does the normal check
1372 	 * "ac_enaddr == ether_dhost". The check sequence is a bit
1373 	 * different from OpenBSD, so we jump over as few code as
1374 	 * possible, to catch _all_ sanity checks. This needs
1375 	 * evaluation, to see if the carp ether_dhost values break any
1376 	 * of these checks!
1377 	 */
1378 	if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost))
1379 		goto post_stats;
1380 #endif
1381 
1382 	/*
1383 	 * Discard packet if upper layers shouldn't see it because
1384 	 * it was unicast to a different Ethernet address.  If the
1385 	 * driver is working properly, then this situation can only
1386 	 * happen when the interface is in promiscuous mode.
1387 	 */
1388 	if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
1389 	    (eh->ether_dhost[0] & 1) == 0 &&
1390 	    bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
1391 		m_freem(m);
1392 		return;
1393 	}
1394 
1395 post_stats:
1396 	if (IPFW_LOADED && ether_ipfw != 0) {
1397 		struct ether_header save_eh = *eh;
1398 
1399 		/* XXX old crufty stuff, needs to be removed */
1400 		m_adj(m, sizeof(struct ether_header));
1401 
1402 		if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
1403 			m_freem(m);
1404 			return;
1405 		}
1406 
1407 		ether_restore_header(&m, eh, &save_eh);
1408 		if (m == NULL)
1409 			return;
1410 		eh = mtod(m, struct ether_header *);
1411 	}
1412 
1413 	ether_type = ntohs(eh->ether_type);
1414 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1415 
1416 	if (m->m_flags & M_VLANTAG) {
1417 		if (vlan_input2_p != NULL) {
1418 			vlan_input2_p(m);
1419 		} else {
1420 			m->m_pkthdr.rcvif->if_noproto++;
1421 			m_freem(m);
1422 		}
1423 		return;
1424 	}
1425 
1426 	m_adj(m, sizeof(struct ether_header));
1427 	redispatch = 0;
1428 
1429 	switch (ether_type) {
1430 #ifdef INET
1431 	case ETHERTYPE_IP:
1432 #ifdef notyet
1433 		if (ipflow_fastforward(m, ifp->if_serializer))
1434 			return;
1435 #endif
1436 		isr = NETISR_IP;
1437 		break;
1438 
1439 	case ETHERTYPE_ARP:
1440 		if (ifp->if_flags & IFF_NOARP) {
1441 			/* Discard packet if ARP is disabled on interface */
1442 			m_freem(m);
1443 			return;
1444 		}
1445 		isr = NETISR_ARP;
1446 		break;
1447 #endif
1448 
1449 #ifdef INET6
1450 	case ETHERTYPE_IPV6:
1451 		isr = NETISR_IPV6;
1452 		break;
1453 #endif
1454 
1455 #ifdef IPX
1456 	case ETHERTYPE_IPX:
1457 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
1458 			return;
1459 		isr = NETISR_IPX;
1460 		break;
1461 #endif
1462 
1463 #ifdef NS
1464 	case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
1465 		isr = NETISR_NS;
1466 		break;
1467 
1468 #endif
1469 
1470 #ifdef NETATALK
1471 	case ETHERTYPE_AT:
1472 		isr = NETISR_ATALK1;
1473 		break;
1474 	case ETHERTYPE_AARP:
1475 		isr = NETISR_AARP;
1476 		break;
1477 #endif
1478 
1479 	default:
1480 		/*
1481 		 * The accurate msgport is not determined before
1482 		 * we reach here, so redo the dispatching
1483 		 */
1484 		redispatch = 1;
1485 #ifdef IPX
1486 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
1487 			return;
1488 #endif
1489 #ifdef NS
1490 		checksum = mtod(m, ushort *);
1491 		/* Novell 802.3 */
1492 		if ((ether_type <= ETHERMTU) &&
1493 		    ((*checksum == 0xffff) || (*checksum == 0xE0E0))) {
1494 			if (*checksum == 0xE0E0) {
1495 				m->m_pkthdr.len -= 3;
1496 				m->m_len -= 3;
1497 				m->m_data += 3;
1498 			}
1499 			isr = NETISR_NS;
1500 			break;
1501 		}
1502 #endif
1503 #ifdef NETATALK
1504 		if (ether_type > ETHERMTU)
1505 			goto dropanyway;
1506 		l = mtod(m, struct llc *);
1507 		if (l->llc_dsap == LLC_SNAP_LSAP &&
1508 		    l->llc_ssap == LLC_SNAP_LSAP &&
1509 		    l->llc_control == LLC_UI) {
1510 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
1511 				 sizeof at_org_code) == 0 &&
1512 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
1513 				m_adj(m, sizeof(struct llc));
1514 				isr = NETISR_ATALK2;
1515 				break;
1516 			}
1517 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
1518 				 sizeof aarp_org_code) == 0 &&
1519 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
1520 				m_adj(m, sizeof(struct llc));
1521 				isr = NETISR_AARP;
1522 				break;
1523 			}
1524 		}
1525 dropanyway:
1526 #endif
1527 		if (ng_ether_input_orphan_p != NULL)
1528 			ng_ether_input_orphan_p(ifp, m, eh);
1529 		else
1530 			m_freem(m);
1531 		return;
1532 	}
1533 
1534 	if (!redispatch)
1535 		netisr_run(isr, m);
1536 	else
1537 		netisr_dispatch(isr, m);
1538 }
1539 
1540 void
1541 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1542 {
1543 	if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1544 		/*
1545 		 * Receiving interface's flags are changed, when this
1546 		 * packet is waiting for processing; discard it.
1547 		 */
1548 		m_freem(m);
1549 		return;
1550 	}
1551 
1552 	/*
1553 	 * Tap the packet off here for a bridge.  bridge_input()
1554 	 * will return NULL if it has consumed the packet, otherwise
1555 	 * it gets processed as normal.  Note that bridge_input()
1556 	 * will always return the original packet if we need to
1557 	 * process it locally.
1558 	 */
1559 	if (ifp->if_bridge) {
1560 		KASSERT(bridge_input_p != NULL,
1561 			("%s: if_bridge not loaded!", __func__));
1562 
1563 		if(m->m_flags & M_PROTO1) {
1564 			m->m_flags &= ~M_PROTO1;
1565 		} else {
1566 			/* clear M_PROMISC, in case the packets comes from a vlan */
1567 			/* m->m_flags &= ~M_PROMISC; */
1568 			m = bridge_input_p(ifp, m);
1569 			if (m == NULL)
1570 				return;
1571 
1572 			KASSERT(ifp == m->m_pkthdr.rcvif,
1573 				("bridge_input_p changed rcvif\n"));
1574 		}
1575 	}
1576 
1577 	/* Handle ng_ether(4) processing, if any */
1578 	if (ng_ether_input_p != NULL) {
1579 		ng_ether_input_p(ifp, &m);
1580 		if (m == NULL)
1581 			return;
1582 	}
1583 
1584 	/* Continue with upper layer processing */
1585 	ether_demux_oncpu(ifp, m);
1586 }
1587 
1588 static void
1589 ether_input_handler(struct netmsg *nmsg)
1590 {
1591 	struct netmsg_packet *nmp = (struct netmsg_packet *)nmsg;
1592 	struct ifnet *ifp;
1593 	struct mbuf *m;
1594 
1595 	m = nmp->nm_packet;
1596 	M_ASSERTPKTHDR(m);
1597 	ifp = m->m_pkthdr.rcvif;
1598 
1599 	ether_input_oncpu(ifp, m);
1600 }
1601 
1602 static __inline void
1603 ether_init_netpacket(int num, struct mbuf *m)
1604 {
1605 	struct netmsg_packet *pmsg;
1606 
1607 	pmsg = &m->m_hdr.mh_netmsg;
1608 	netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport, 0,
1609 		    ether_input_handler);
1610 	pmsg->nm_packet = m;
1611 	pmsg->nm_netmsg.nm_lmsg.u.ms_result = num;
1612 }
1613 
1614 static __inline struct lwkt_port *
1615 ether_mport(int num, struct mbuf **m0)
1616 {
1617 	struct lwkt_port *port;
1618 	struct mbuf *m = *m0;
1619 
1620 	if (num == NETISR_MAX) {
1621 		/*
1622 		 * All packets whose target msgports can't be
1623 		 * determined here are dispatched to netisr0,
1624 		 * where further dispatching may happen.
1625 		 */
1626 		return cpu_portfn(0);
1627 	}
1628 
1629 	port = netisr_find_port(num, &m);
1630 	if (port == NULL)
1631 		return NULL;
1632 
1633 	*m0 = m;
1634 	return port;
1635 }
1636 
1637 void
1638 ether_input_chain2(struct ifnet *ifp, struct mbuf *m, struct mbuf_chain *chain)
1639 {
1640 	struct ether_header *eh, *save_eh, save_eh0;
1641 	struct lwkt_port *port;
1642 	uint16_t ether_type;
1643 	int isr;
1644 
1645 	ASSERT_SERIALIZED(ifp->if_serializer);
1646 	M_ASSERTPKTHDR(m);
1647 
1648 	/* Discard packet if interface is not up */
1649 	if (!(ifp->if_flags & IFF_UP)) {
1650 		m_freem(m);
1651 		return;
1652 	}
1653 
1654 	if (m->m_len < sizeof(struct ether_header)) {
1655 		/* XXX error in the caller. */
1656 		m_freem(m);
1657 		return;
1658 	}
1659 	eh = mtod(m, struct ether_header *);
1660 
1661 	m->m_pkthdr.rcvif = ifp;
1662 
1663 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1664 		if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1665 			 ifp->if_addrlen) == 0)
1666 			m->m_flags |= M_BCAST;
1667 		else
1668 			m->m_flags |= M_MCAST;
1669 		ifp->if_imcasts++;
1670 	}
1671 
1672 	ETHER_BPF_MTAP(ifp, m);
1673 
1674 	ifp->if_ibytes += m->m_pkthdr.len;
1675 
1676 	if (ifp->if_flags & IFF_MONITOR) {
1677 		/*
1678 		 * Interface marked for monitoring; discard packet.
1679 		 */
1680 		m_freem(m);
1681 		return;
1682 	}
1683 
1684 	if (ntohs(eh->ether_type) == ETHERTYPE_VLAN &&
1685 	    (m->m_flags & M_VLANTAG) == 0) {
1686 		/*
1687 		 * Extract vlan tag if hardware does not do it for us
1688 		 */
1689 		vlan_ether_decap(&m);
1690 		if (m == NULL)
1691 			return;
1692 		eh = mtod(m, struct ether_header *);
1693 	}
1694 	ether_type = ntohs(eh->ether_type);
1695 
1696 	if ((m->m_flags & M_VLANTAG) && ether_type == ETHERTYPE_VLAN) {
1697 		/*
1698 		 * To prevent possible dangerous recursion,
1699 		 * we don't do vlan-in-vlan
1700 		 */
1701 		ifp->if_noproto++;
1702 		m_freem(m);
1703 		return;
1704 	}
1705 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1706 
1707 	/*
1708 	 * Map ether type to netisr id.
1709 	 */
1710 	switch (ether_type) {
1711 #ifdef INET
1712 	case ETHERTYPE_IP:
1713 		isr = NETISR_IP;
1714 		break;
1715 
1716 	case ETHERTYPE_ARP:
1717 		isr = NETISR_ARP;
1718 		break;
1719 #endif
1720 
1721 #ifdef INET6
1722 	case ETHERTYPE_IPV6:
1723 		isr = NETISR_IPV6;
1724 		break;
1725 #endif
1726 
1727 #ifdef IPX
1728 	case ETHERTYPE_IPX:
1729 		isr = NETISR_IPX;
1730 		break;
1731 #endif
1732 
1733 #ifdef NS
1734 	case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
1735 		isr = NETISR_NS;
1736 		break;
1737 #endif
1738 
1739 #ifdef NETATALK
1740 	case ETHERTYPE_AT:
1741 		isr = NETISR_ATALK1;
1742 		break;
1743 	case ETHERTYPE_AARP:
1744 		isr = NETISR_AARP;
1745 		break;
1746 #endif
1747 
1748 	default:
1749 		/*
1750 		 * NETISR_MAX is an invalid value; it is chosen to let
1751 		 * ether_mport() know that we are not able to decide
1752 		 * this packet's msgport here.
1753 		 */
1754 		isr = NETISR_MAX;
1755 		break;
1756 	}
1757 
1758 	/*
1759 	 * If the packet is in contiguous memory, following
1760 	 * m_adj() could ensure that the hidden ether header
1761 	 * will not be destroyed, else we will have to save
1762 	 * the ether header for the later restoration.
1763 	 */
1764 	if (m->m_pkthdr.len != m->m_len) {
1765 		save_eh0 = *eh;
1766 		save_eh = &save_eh0;
1767 	} else {
1768 		save_eh = NULL;
1769 	}
1770 
1771 	/*
1772 	 * Temporarily remove ether header; ether_mport()
1773 	 * expects a packet without ether header.
1774 	 */
1775 	m_adj(m, sizeof(struct ether_header));
1776 
1777 	/*
1778 	 * Find the packet's target msgport.
1779 	 */
1780 	port = ether_mport(isr, &m);
1781 	if (port == NULL) {
1782 		KKASSERT(m == NULL);
1783 		return;
1784 	}
1785 
1786 	/*
1787 	 * Restore ether header.
1788 	 */
1789 	if (save_eh != NULL) {
1790 		ether_restore_header(&m, eh, save_eh);
1791 		if (m == NULL)
1792 			return;
1793 	} else {
1794 		m->m_data -= ETHER_HDR_LEN;
1795 		m->m_len += ETHER_HDR_LEN;
1796 		m->m_pkthdr.len += ETHER_HDR_LEN;
1797 	}
1798 
1799 	/*
1800 	 * Initialize mbuf's netmsg packet _after_ possible
1801 	 * ether header restoration, else the initialized
1802 	 * netmsg packet may be lost during ether header
1803 	 * restoration.
1804 	 */
1805 	ether_init_netpacket(isr, m);
1806 
1807 #ifdef ETHER_INPUT_CHAIN
1808 	if (chain != NULL) {
1809 		struct mbuf_chain *c;
1810 		int cpuid;
1811 
1812 		m->m_pkthdr.header = port; /* XXX */
1813 		cpuid = port->mpu_td->td_gd->gd_cpuid;
1814 
1815 		c = &chain[cpuid];
1816 		if (c->mc_head == NULL) {
1817 			c->mc_head = c->mc_tail = m;
1818 		} else {
1819 			c->mc_tail->m_nextpkt = m;
1820 			c->mc_tail = m;
1821 		}
1822 		m->m_nextpkt = NULL;
1823 	} else
1824 #endif	/* ETHER_INPUT_CHAIN */
1825 		lwkt_sendmsg(port, &m->m_hdr.mh_netmsg.nm_netmsg.nm_lmsg);
1826 }
1827 
1828 #endif	/* ETHER_INPUT2 */
1829