xref: /dflybsd-src/sys/net/if_ethersubr.c (revision 92db3805bd7a1f247cdbdcfd50157ca600010b24)
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 1.90 2008/09/20 10:53:16 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_mpls.h"
43 #include "opt_netgraph.h"
44 #include "opt_carp.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/ktr.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 void	(*vlan_input_p)(struct mbuf *);
128 
129 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
130 			struct rtentry *);
131 static void ether_restore_header(struct mbuf **, const struct ether_header *,
132 				 const struct ether_header *);
133 
134 /*
135  * if_bridge support
136  */
137 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
138 int (*bridge_output_p)(struct ifnet *, struct mbuf *);
139 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
140 
141 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
142 			      struct sockaddr *);
143 
144 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
145 	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
146 };
147 
148 #define gotoerr(e) do { error = (e); goto bad; } while (0)
149 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
150 
151 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
152 				struct ip_fw **rule,
153 				const struct ether_header *eh);
154 
155 static int ether_ipfw;
156 static u_int ether_restore_hdr;
157 static u_int ether_prepend_hdr;
158 
159 SYSCTL_DECL(_net_link);
160 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
161 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
162 	   &ether_ipfw, 0, "Pass ether pkts through firewall");
163 SYSCTL_UINT(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW,
164 	    &ether_restore_hdr, 0, "# of ether header restoration");
165 SYSCTL_UINT(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW,
166 	    &ether_prepend_hdr, 0,
167 	    "# of ether header restoration which prepends mbuf");
168 
169 #define ETHER_KTR_STR		"ifp=%p"
170 #define ETHER_KTR_ARG_SIZE	(sizeof(void *))
171 #ifndef KTR_ETHERNET
172 #define KTR_ETHERNET		KTR_ALL
173 #endif
174 KTR_INFO_MASTER(ether);
175 KTR_INFO(KTR_ETHERNET, ether, chain_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARG_SIZE);
176 KTR_INFO(KTR_ETHERNET, ether, chain_end, 1, ETHER_KTR_STR, ETHER_KTR_ARG_SIZE);
177 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARG_SIZE);
178 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARG_SIZE);
179 #define logether(name, arg)	KTR_LOG(ether_ ## name, arg)
180 
181 /*
182  * Ethernet output routine.
183  * Encapsulate a packet of type family for the local net.
184  * Use trailer local net encapsulation if enough data in first
185  * packet leaves a multiple of 512 bytes of data in remainder.
186  * Assumes that ifp is actually pointer to arpcom structure.
187  */
188 static int
189 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
190 	     struct rtentry *rt)
191 {
192 	struct ether_header *eh, *deh;
193 	u_char *edst;
194 	int loop_copy = 0;
195 	int hlen = ETHER_HDR_LEN;	/* link layer header length */
196 	struct arpcom *ac = IFP2AC(ifp);
197 	int error;
198 
199 	ASSERT_NOT_SERIALIZED(ifp->if_serializer);
200 
201 	if (ifp->if_flags & IFF_MONITOR)
202 		gotoerr(ENETDOWN);
203 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
204 		gotoerr(ENETDOWN);
205 
206 	M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT);
207 	if (m == NULL)
208 		return (ENOBUFS);
209 	eh = mtod(m, struct ether_header *);
210 	edst = eh->ether_dhost;
211 
212 	/*
213 	 * Fill in the destination ethernet address and frame type.
214 	 */
215 	switch (dst->sa_family) {
216 #ifdef INET
217 	case AF_INET:
218 		if (!arpresolve(ifp, rt, m, dst, edst))
219 			return (0);	/* if not yet resolved */
220 #ifdef MPLS
221 		if (m->m_flags & M_MPLSLABELED)
222 			eh->ether_type = htons(ETHERTYPE_MPLS);
223 		else
224 #endif
225 			eh->ether_type = htons(ETHERTYPE_IP);
226 		break;
227 #endif
228 #ifdef INET6
229 	case AF_INET6:
230 		if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst))
231 			return (0);		/* Something bad happenned. */
232 		eh->ether_type = htons(ETHERTYPE_IPV6);
233 		break;
234 #endif
235 #ifdef IPX
236 	case AF_IPX:
237 		if (ef_outputp != NULL) {
238 			error = ef_outputp(ifp, &m, dst, &eh->ether_type,
239 					   &hlen);
240 			if (error)
241 				goto bad;
242 		} else {
243 			eh->ether_type = htons(ETHERTYPE_IPX);
244 			bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
245 			      edst, ETHER_ADDR_LEN);
246 		}
247 		break;
248 #endif
249 #ifdef NETATALK
250 	case AF_APPLETALK: {
251 		struct at_ifaddr *aa;
252 
253 		if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL) {
254 			error = 0;	/* XXX */
255 			goto bad;
256 		}
257 		/*
258 		 * In the phase 2 case, need to prepend an mbuf for
259 		 * the llc header.  Since we must preserve the value
260 		 * of m, which is passed to us by value, we m_copy()
261 		 * the first mbuf, and use it for our llc header.
262 		 */
263 		if (aa->aa_flags & AFA_PHASE2) {
264 			struct llc llc;
265 
266 			M_PREPEND(m, sizeof(struct llc), MB_DONTWAIT);
267 			eh = mtod(m, struct ether_header *);
268 			edst = eh->ether_dhost;
269 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
270 			llc.llc_control = LLC_UI;
271 			bcopy(at_org_code, llc.llc_snap_org_code,
272 			      sizeof at_org_code);
273 			llc.llc_snap_ether_type = htons(ETHERTYPE_AT);
274 			bcopy(&llc,
275 			      mtod(m, caddr_t) + sizeof(struct ether_header),
276 			      sizeof(struct llc));
277 			eh->ether_type = htons(m->m_pkthdr.len);
278 			hlen = sizeof(struct llc) + ETHER_HDR_LEN;
279 		} else {
280 			eh->ether_type = htons(ETHERTYPE_AT);
281 		}
282 		if (!aarpresolve(ac, m, (struct sockaddr_at *)dst, edst))
283 			return (0);
284 		break;
285 	  }
286 #endif
287 #ifdef NS
288 	case AF_NS:
289 		switch(ns_nettype) {
290 		default:
291 		case 0x8137:	/* Novell Ethernet_II Ethernet TYPE II */
292 			eh->ether_type = 0x8137;
293 			break;
294 		case 0x0:	/* Novell 802.3 */
295 			eh->ether_type = htons(m->m_pkthdr.len);
296 			break;
297 		case 0xe0e0:	/* Novell 802.2 and Token-Ring */
298 			M_PREPEND(m, 3, MB_DONTWAIT);
299 			eh = mtod(m, struct ether_header *);
300 			edst = eh->ether_dhost;
301 			eh->ether_type = htons(m->m_pkthdr.len);
302 			cp = mtod(m, u_char *) + sizeof(struct ether_header);
303 			*cp++ = 0xE0;
304 			*cp++ = 0xE0;
305 			*cp++ = 0x03;
306 			break;
307 		}
308 		bcopy(&(((struct sockaddr_ns *)dst)->sns_addr.x_host), edst,
309 		      ETHER_ADDR_LEN);
310 		/*
311 		 * XXX if ns_thishost is the same as the node's ethernet
312 		 * address then just the default code will catch this anyhow.
313 		 * So I'm not sure if this next clause should be here at all?
314 		 * [JRE]
315 		 */
316 		if (bcmp(edst, &ns_thishost, ETHER_ADDR_LEN) == 0) {
317 			m->m_pkthdr.rcvif = ifp;
318 			netisr_dispatch(NETISR_NS, m);
319 			return (error);
320 		}
321 		if (bcmp(edst, &ns_broadhost, ETHER_ADDR_LEN) == 0)
322 			m->m_flags |= M_BCAST;
323 		break;
324 #endif
325 	case pseudo_AF_HDRCMPLT:
326 	case AF_UNSPEC:
327 		loop_copy = -1; /* if this is for us, don't do it */
328 		deh = (struct ether_header *)dst->sa_data;
329 		memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
330 		eh->ether_type = deh->ether_type;
331 		break;
332 
333 	default:
334 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
335 		gotoerr(EAFNOSUPPORT);
336 	}
337 
338 	if (dst->sa_family == pseudo_AF_HDRCMPLT)	/* unlikely */
339 		memcpy(eh->ether_shost,
340 		       ((struct ether_header *)dst->sa_data)->ether_shost,
341 		       ETHER_ADDR_LEN);
342 	else
343 		memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
344 
345 	/*
346 	 * Bridges require special output handling.
347 	 */
348 	if (ifp->if_bridge) {
349 		KASSERT(bridge_output_p != NULL,
350 			("%s: if_bridge not loaded!", __func__));
351 		return bridge_output_p(ifp, m);
352 	}
353 
354 	/*
355 	 * If a simplex interface, and the packet is being sent to our
356 	 * Ethernet address or a broadcast address, loopback a copy.
357 	 * XXX To make a simplex device behave exactly like a duplex
358 	 * device, we should copy in the case of sending to our own
359 	 * ethernet address (thus letting the original actually appear
360 	 * on the wire). However, we don't do that here for security
361 	 * reasons and compatibility with the original behavior.
362 	 */
363 	if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
364 		int csum_flags = 0;
365 
366 		if (m->m_pkthdr.csum_flags & CSUM_IP)
367 			csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
368 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
369 			csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
370 		if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
371 			struct mbuf *n;
372 
373 			if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) {
374 				n->m_pkthdr.csum_flags |= csum_flags;
375 				if (csum_flags & CSUM_DATA_VALID)
376 					n->m_pkthdr.csum_data = 0xffff;
377 				if_simloop(ifp, n, dst->sa_family, hlen);
378 			} else
379 				ifp->if_iqdrops++;
380 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
381 				ETHER_ADDR_LEN) == 0) {
382 			m->m_pkthdr.csum_flags |= csum_flags;
383 			if (csum_flags & CSUM_DATA_VALID)
384 				m->m_pkthdr.csum_data = 0xffff;
385 			if_simloop(ifp, m, dst->sa_family, hlen);
386 			return (0);	/* XXX */
387 		}
388 	}
389 
390 #ifdef CARP
391 	if (ifp->if_carp && (error = carp_output(ifp, m, dst, NULL)))
392 		goto bad;
393 #endif
394 
395 
396 	/* Handle ng_ether(4) processing, if any */
397 	if (ng_ether_output_p != NULL) {
398 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0)
399 			goto bad;
400 		if (m == NULL)
401 			return (0);
402 	}
403 
404 	/* Continue with link-layer output */
405 	return ether_output_frame(ifp, m);
406 
407 bad:
408 	m_freem(m);
409 	return (error);
410 }
411 
412 /*
413  * Ethernet link layer output routine to send a raw frame to the device.
414  *
415  * This assumes that the 14 byte Ethernet header is present and contiguous
416  * in the first mbuf.
417  */
418 int
419 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
420 {
421 	struct ip_fw *rule = NULL;
422 	int error = 0;
423 	struct altq_pktattr pktattr;
424 
425 	ASSERT_NOT_SERIALIZED(ifp->if_serializer);
426 
427 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
428 		struct m_tag *mtag;
429 
430 		/* Extract info from dummynet tag */
431 		mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
432 		KKASSERT(mtag != NULL);
433 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
434 		KKASSERT(rule != NULL);
435 
436 		m_tag_delete(m, mtag);
437 		m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
438 	}
439 
440 	if (ifq_is_enabled(&ifp->if_snd))
441 		altq_etherclassify(&ifp->if_snd, m, &pktattr);
442 	crit_enter();
443 	if (IPFW_LOADED && ether_ipfw != 0) {
444 		struct ether_header save_eh, *eh;
445 
446 		eh = mtod(m, struct ether_header *);
447 		save_eh = *eh;
448 		m_adj(m, ETHER_HDR_LEN);
449 		if (!ether_ipfw_chk(&m, ifp, &rule, eh)) {
450 			crit_exit();
451 			if (m != NULL) {
452 				m_freem(m);
453 				return ENOBUFS; /* pkt dropped */
454 			} else
455 				return 0;	/* consumed e.g. in a pipe */
456 		}
457 
458 		/* packet was ok, restore the ethernet header */
459 		ether_restore_header(&m, eh, &save_eh);
460 		if (m == NULL) {
461 			crit_exit();
462 			return ENOBUFS;
463 		}
464 	}
465 	crit_exit();
466 
467 	/*
468 	 * Queue message on interface, update output statistics if
469 	 * successful, and start output if interface not yet active.
470 	 */
471 	error = ifq_dispatch(ifp, m, &pktattr);
472 	return (error);
473 }
474 
475 /*
476  * ipfw processing for ethernet packets (in and out).
477  * The second parameter is NULL from ether_demux(), and ifp from
478  * ether_output_frame().
479  */
480 static boolean_t
481 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule,
482 	       const struct ether_header *eh)
483 {
484 	struct ether_header save_eh = *eh;	/* might be a ptr in *m0 */
485 	struct ip_fw_args args;
486 	struct m_tag *mtag;
487 	struct mbuf *m;
488 	int i;
489 
490 	if (*rule != NULL && fw_one_pass)
491 		return TRUE; /* dummynet packet, already partially processed */
492 
493 	/*
494 	 * I need some amount of data to be contiguous.
495 	 */
496 	i = min((*m0)->m_pkthdr.len, max_protohdr);
497 	if ((*m0)->m_len < i) {
498 		*m0 = m_pullup(*m0, i);
499 		if (*m0 == NULL)
500 			return FALSE;
501 	}
502 
503 	/*
504 	 * Clean up tags
505 	 */
506 	if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
507 		m_tag_delete(*m0, mtag);
508 	if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
509 		mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL);
510 		KKASSERT(mtag != NULL);
511 		m_tag_delete(*m0, mtag);
512 		(*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED;
513 	}
514 
515 	args.m = *m0;		/* the packet we are looking at		*/
516 	args.oif = dst;		/* destination, if any			*/
517 	args.rule = *rule;	/* matching rule to restart		*/
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 (*m0 == NULL)
524 		return FALSE;
525 
526 	switch (i) {
527 	case IP_FW_PASS:
528 		return TRUE;
529 
530 	case IP_FW_DIVERT:
531 	case IP_FW_TEE:
532 	case IP_FW_DENY:
533 		/*
534 		 * XXX at some point add support for divert/forward actions.
535 		 * If none of the above matches, we have to drop the pkt.
536 		 */
537 		return FALSE;
538 
539 	case IP_FW_DUMMYNET:
540 		/*
541 		 * Pass the pkt to dummynet, which consumes it.
542 		 */
543 		m = *m0;	/* pass the original to dummynet */
544 		*m0 = NULL;	/* and nothing back to the caller */
545 
546 		ether_restore_header(&m, eh, &save_eh);
547 		if (m == NULL)
548 			return FALSE;
549 
550 		ip_fw_dn_io_ptr(m, args.cookie,
551 				dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
552 		ip_dn_queue(m);
553 		return FALSE;
554 
555 	default:
556 		panic("unknown ipfw return value: %d\n", i);
557 	}
558 }
559 
560 static void
561 ether_input(struct ifnet *ifp, struct mbuf *m)
562 {
563 	ether_input_chain(ifp, m, NULL);
564 }
565 
566 /*
567  * Perform common duties while attaching to interface list
568  */
569 void
570 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer)
571 {
572 	ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
573 			   serializer);
574 }
575 
576 void
577 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen,
578 		   lwkt_serialize_t serializer)
579 {
580 	struct sockaddr_dl *sdl;
581 
582 	ifp->if_type = IFT_ETHER;
583 	ifp->if_addrlen = ETHER_ADDR_LEN;
584 	ifp->if_hdrlen = ETHER_HDR_LEN;
585 	if_attach(ifp, serializer);
586 	ifp->if_mtu = ETHERMTU;
587 	if (ifp->if_baudrate == 0)
588 		ifp->if_baudrate = 10000000;
589 	ifp->if_output = ether_output;
590 	ifp->if_input = ether_input;
591 	ifp->if_resolvemulti = ether_resolvemulti;
592 	ifp->if_broadcastaddr = etherbroadcastaddr;
593 	sdl = IF_LLSOCKADDR(ifp);
594 	sdl->sdl_type = IFT_ETHER;
595 	sdl->sdl_alen = ifp->if_addrlen;
596 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
597 	/*
598 	 * XXX Keep the current drivers happy.
599 	 * XXX Remove once all drivers have been cleaned up
600 	 */
601 	if (lla != IFP2AC(ifp)->ac_enaddr)
602 		bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
603 	bpfattach(ifp, dlt, hdrlen);
604 	if (ng_ether_attach_p != NULL)
605 		(*ng_ether_attach_p)(ifp);
606 
607 	if_printf(ifp, "MAC address: %6D\n", lla, ":");
608 }
609 
610 /*
611  * Perform common duties while detaching an Ethernet interface
612  */
613 void
614 ether_ifdetach(struct ifnet *ifp)
615 {
616 	if_down(ifp);
617 
618 	if (ng_ether_detach_p != NULL)
619 		(*ng_ether_detach_p)(ifp);
620 	bpfdetach(ifp);
621 	if_detach(ifp);
622 }
623 
624 int
625 ether_ioctl(struct ifnet *ifp, int command, caddr_t data)
626 {
627 	struct ifaddr *ifa = (struct ifaddr *) data;
628 	struct ifreq *ifr = (struct ifreq *) data;
629 	int error = 0;
630 
631 #define IF_INIT(ifp) \
632 do { \
633 	if (((ifp)->if_flags & IFF_UP) == 0) { \
634 		(ifp)->if_flags |= IFF_UP; \
635 		(ifp)->if_init((ifp)->if_softc); \
636 	} \
637 } while (0)
638 
639 	ASSERT_SERIALIZED(ifp->if_serializer);
640 
641 	switch (command) {
642 	case SIOCSIFADDR:
643 		switch (ifa->ifa_addr->sa_family) {
644 #ifdef INET
645 		case AF_INET:
646 			IF_INIT(ifp);	/* before arpwhohas */
647 			arp_ifinit(ifp, ifa);
648 			break;
649 #endif
650 #ifdef IPX
651 		/*
652 		 * XXX - This code is probably wrong
653 		 */
654 		case AF_IPX:
655 			{
656 			struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr;
657 			struct arpcom *ac = IFP2AC(ifp);
658 
659 			if (ipx_nullhost(*ina))
660 				ina->x_host = *(union ipx_host *) ac->ac_enaddr;
661 			else
662 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
663 				      sizeof ac->ac_enaddr);
664 
665 			IF_INIT(ifp);	/* Set new address. */
666 			break;
667 			}
668 #endif
669 #ifdef NS
670 		/*
671 		 * XXX - This code is probably wrong
672 		 */
673 		case AF_NS:
674 		{
675 			struct ns_addr *ina = &(IA_SNS(ifa)->sns_addr);
676 			struct arpcom *ac = IFP2AC(ifp);
677 
678 			if (ns_nullhost(*ina))
679 				ina->x_host = *(union ns_host *)(ac->ac_enaddr);
680 			else
681 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
682 				      sizeof ac->ac_enaddr);
683 
684 			/*
685 			 * Set new address
686 			 */
687 			IF_INIT(ifp);
688 			break;
689 		}
690 #endif
691 		default:
692 			IF_INIT(ifp);
693 			break;
694 		}
695 		break;
696 
697 	case SIOCGIFADDR:
698 		bcopy(IFP2AC(ifp)->ac_enaddr,
699 		      ((struct sockaddr *)ifr->ifr_data)->sa_data,
700 		      ETHER_ADDR_LEN);
701 		break;
702 
703 	case SIOCSIFMTU:
704 		/*
705 		 * Set the interface MTU.
706 		 */
707 		if (ifr->ifr_mtu > ETHERMTU) {
708 			error = EINVAL;
709 		} else {
710 			ifp->if_mtu = ifr->ifr_mtu;
711 		}
712 		break;
713 	default:
714 		error = EINVAL;
715 		break;
716 	}
717 	return (error);
718 
719 #undef IF_INIT
720 }
721 
722 int
723 ether_resolvemulti(
724 	struct ifnet *ifp,
725 	struct sockaddr **llsa,
726 	struct sockaddr *sa)
727 {
728 	struct sockaddr_dl *sdl;
729 	struct sockaddr_in *sin;
730 #ifdef INET6
731 	struct sockaddr_in6 *sin6;
732 #endif
733 	u_char *e_addr;
734 
735 	switch(sa->sa_family) {
736 	case AF_LINK:
737 		/*
738 		 * No mapping needed. Just check that it's a valid MC address.
739 		 */
740 		sdl = (struct sockaddr_dl *)sa;
741 		e_addr = LLADDR(sdl);
742 		if ((e_addr[0] & 1) != 1)
743 			return EADDRNOTAVAIL;
744 		*llsa = 0;
745 		return 0;
746 
747 #ifdef INET
748 	case AF_INET:
749 		sin = (struct sockaddr_in *)sa;
750 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
751 			return EADDRNOTAVAIL;
752 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
753 		       M_WAITOK | M_ZERO);
754 		sdl->sdl_len = sizeof *sdl;
755 		sdl->sdl_family = AF_LINK;
756 		sdl->sdl_index = ifp->if_index;
757 		sdl->sdl_type = IFT_ETHER;
758 		sdl->sdl_alen = ETHER_ADDR_LEN;
759 		e_addr = LLADDR(sdl);
760 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
761 		*llsa = (struct sockaddr *)sdl;
762 		return 0;
763 #endif
764 #ifdef INET6
765 	case AF_INET6:
766 		sin6 = (struct sockaddr_in6 *)sa;
767 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
768 			/*
769 			 * An IP6 address of 0 means listen to all
770 			 * of the Ethernet multicast address used for IP6.
771 			 * (This is used for multicast routers.)
772 			 */
773 			ifp->if_flags |= IFF_ALLMULTI;
774 			*llsa = 0;
775 			return 0;
776 		}
777 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
778 			return EADDRNOTAVAIL;
779 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
780 		       M_WAITOK | M_ZERO);
781 		sdl->sdl_len = sizeof *sdl;
782 		sdl->sdl_family = AF_LINK;
783 		sdl->sdl_index = ifp->if_index;
784 		sdl->sdl_type = IFT_ETHER;
785 		sdl->sdl_alen = ETHER_ADDR_LEN;
786 		e_addr = LLADDR(sdl);
787 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
788 		*llsa = (struct sockaddr *)sdl;
789 		return 0;
790 #endif
791 
792 	default:
793 		/*
794 		 * Well, the text isn't quite right, but it's the name
795 		 * that counts...
796 		 */
797 		return EAFNOSUPPORT;
798 	}
799 }
800 
801 #if 0
802 /*
803  * This is for reference.  We have a table-driven version
804  * of the little-endian crc32 generator, which is faster
805  * than the double-loop.
806  */
807 uint32_t
808 ether_crc32_le(const uint8_t *buf, size_t len)
809 {
810 	uint32_t c, crc, carry;
811 	size_t i, j;
812 
813 	crc = 0xffffffffU;	/* initial value */
814 
815 	for (i = 0; i < len; i++) {
816 		c = buf[i];
817 		for (j = 0; j < 8; j++) {
818 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
819 			crc >>= 1;
820 			c >>= 1;
821 			if (carry)
822 				crc = (crc ^ ETHER_CRC_POLY_LE);
823 		}
824 	}
825 
826 	return (crc);
827 }
828 #else
829 uint32_t
830 ether_crc32_le(const uint8_t *buf, size_t len)
831 {
832 	static const uint32_t crctab[] = {
833 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
834 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
835 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
836 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
837 	};
838 	uint32_t crc;
839 	size_t i;
840 
841 	crc = 0xffffffffU;	/* initial value */
842 
843 	for (i = 0; i < len; i++) {
844 		crc ^= buf[i];
845 		crc = (crc >> 4) ^ crctab[crc & 0xf];
846 		crc = (crc >> 4) ^ crctab[crc & 0xf];
847 	}
848 
849 	return (crc);
850 }
851 #endif
852 
853 uint32_t
854 ether_crc32_be(const uint8_t *buf, size_t len)
855 {
856 	uint32_t c, crc, carry;
857 	size_t i, j;
858 
859 	crc = 0xffffffffU;	/* initial value */
860 
861 	for (i = 0; i < len; i++) {
862 		c = buf[i];
863 		for (j = 0; j < 8; j++) {
864 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
865 			crc <<= 1;
866 			c >>= 1;
867 			if (carry)
868 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
869 		}
870 	}
871 
872 	return (crc);
873 }
874 
875 /*
876  * find the size of ethernet header, and call classifier
877  */
878 void
879 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
880 		   struct altq_pktattr *pktattr)
881 {
882 	struct ether_header *eh;
883 	uint16_t ether_type;
884 	int hlen, af, hdrsize;
885 	caddr_t hdr;
886 
887 	hlen = sizeof(struct ether_header);
888 	eh = mtod(m, struct ether_header *);
889 
890 	ether_type = ntohs(eh->ether_type);
891 	if (ether_type < ETHERMTU) {
892 		/* ick! LLC/SNAP */
893 		struct llc *llc = (struct llc *)(eh + 1);
894 		hlen += 8;
895 
896 		if (m->m_len < hlen ||
897 		    llc->llc_dsap != LLC_SNAP_LSAP ||
898 		    llc->llc_ssap != LLC_SNAP_LSAP ||
899 		    llc->llc_control != LLC_UI)
900 			goto bad;  /* not snap! */
901 
902 		ether_type = ntohs(llc->llc_un.type_snap.ether_type);
903 	}
904 
905 	if (ether_type == ETHERTYPE_IP) {
906 		af = AF_INET;
907 		hdrsize = 20;  /* sizeof(struct ip) */
908 #ifdef INET6
909 	} else if (ether_type == ETHERTYPE_IPV6) {
910 		af = AF_INET6;
911 		hdrsize = 40;  /* sizeof(struct ip6_hdr) */
912 #endif
913 	} else
914 		goto bad;
915 
916 	while (m->m_len <= hlen) {
917 		hlen -= m->m_len;
918 		m = m->m_next;
919 	}
920 	hdr = m->m_data + hlen;
921 	if (m->m_len < hlen + hdrsize) {
922 		/*
923 		 * ip header is not in a single mbuf.  this should not
924 		 * happen in the current code.
925 		 * (todo: use m_pulldown in the future)
926 		 */
927 		goto bad;
928 	}
929 	m->m_data += hlen;
930 	m->m_len -= hlen;
931 	ifq_classify(ifq, m, af, pktattr);
932 	m->m_data -= hlen;
933 	m->m_len += hlen;
934 
935 	return;
936 
937 bad:
938 	pktattr->pattr_class = NULL;
939 	pktattr->pattr_hdr = NULL;
940 	pktattr->pattr_af = AF_UNSPEC;
941 }
942 
943 static void
944 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
945 		     const struct ether_header *save_eh)
946 {
947 	struct mbuf *m = *m0;
948 
949 	ether_restore_hdr++;
950 
951 	/*
952 	 * Prepend the header, optimize for the common case of
953 	 * eh pointing into the mbuf.
954 	 */
955 	if ((const void *)(eh + 1) == (void *)m->m_data) {
956 		m->m_data -= ETHER_HDR_LEN;
957 		m->m_len += ETHER_HDR_LEN;
958 		m->m_pkthdr.len += ETHER_HDR_LEN;
959 	} else {
960 		ether_prepend_hdr++;
961 
962 		M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
963 		if (m != NULL) {
964 			bcopy(save_eh, mtod(m, struct ether_header *),
965 			      ETHER_HDR_LEN);
966 		}
967 	}
968 	*m0 = m;
969 }
970 
971 static void
972 ether_input_ipifunc(void *arg)
973 {
974 	struct mbuf *m, *next;
975 	lwkt_port_t port;
976 
977 	m = arg;
978 	do {
979 		next = m->m_nextpkt;
980 		m->m_nextpkt = NULL;
981 
982 		port = m->m_pkthdr.header;
983 		m->m_pkthdr.header = NULL;
984 
985 		lwkt_sendmsg(port,
986 		&m->m_hdr.mh_netmsg.nm_netmsg.nm_lmsg);
987 
988 		m = next;
989 	} while (m != NULL);
990 }
991 
992 void
993 ether_input_dispatch(struct mbuf_chain *chain)
994 {
995 #ifdef SMP
996 	int i;
997 
998 	logether(disp_beg, NULL);
999 	for (i = 0; i < ncpus; ++i) {
1000 		if (chain[i].mc_head != NULL) {
1001 			lwkt_send_ipiq(globaldata_find(i),
1002 			ether_input_ipifunc, chain[i].mc_head);
1003 		}
1004 	}
1005 #else
1006 	logether(disp_beg, NULL);
1007 	if (chain->mc_head != NULL)
1008 		ether_input_ipifunc(chain->mc_head);
1009 #endif
1010 	logether(disp_end, NULL);
1011 }
1012 
1013 void
1014 ether_input_chain_init(struct mbuf_chain *chain)
1015 {
1016 #ifdef SMP
1017 	int i;
1018 
1019 	for (i = 0; i < ncpus; ++i)
1020 		chain[i].mc_head = chain[i].mc_tail = NULL;
1021 #else
1022 	chain->mc_head = chain->mc_tail = NULL;
1023 #endif
1024 }
1025 
1026 /*
1027  * Upper layer processing for a received Ethernet packet.
1028  */
1029 void
1030 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
1031 {
1032 	struct ether_header *eh;
1033 	int isr, redispatch;
1034 	u_short ether_type;
1035 	struct ip_fw *rule = NULL;
1036 #ifdef NETATALK
1037 	struct llc *l;
1038 #endif
1039 
1040 	M_ASSERTPKTHDR(m);
1041 	KASSERT(m->m_len >= ETHER_HDR_LEN,
1042 		("ether header is no contiguous!\n"));
1043 
1044 	eh = mtod(m, struct ether_header *);
1045 
1046 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
1047 		struct m_tag *mtag;
1048 
1049 		/* Extract info from dummynet tag */
1050 		mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
1051 		KKASSERT(mtag != NULL);
1052 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
1053 		KKASSERT(rule != NULL);
1054 
1055 		m_tag_delete(m, mtag);
1056 		m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
1057 
1058 		/* packet is passing the second time */
1059 		goto post_stats;
1060 	}
1061 
1062 #ifdef CARP
1063 	/*
1064 	 * XXX: Okay, we need to call carp_forus() and - if it is for
1065 	 * us jump over code that does the normal check
1066 	 * "ac_enaddr == ether_dhost". The check sequence is a bit
1067 	 * different from OpenBSD, so we jump over as few code as
1068 	 * possible, to catch _all_ sanity checks. This needs
1069 	 * evaluation, to see if the carp ether_dhost values break any
1070 	 * of these checks!
1071 	 */
1072 	if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost))
1073 		goto post_stats;
1074 #endif
1075 
1076 	/*
1077 	 * Discard packet if upper layers shouldn't see it because
1078 	 * it was unicast to a different Ethernet address.  If the
1079 	 * driver is working properly, then this situation can only
1080 	 * happen when the interface is in promiscuous mode.
1081 	 */
1082 	if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
1083 	    (eh->ether_dhost[0] & 1) == 0 &&
1084 	    bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
1085 		m_freem(m);
1086 		return;
1087 	}
1088 
1089 post_stats:
1090 	if (IPFW_LOADED && ether_ipfw != 0) {
1091 		struct ether_header save_eh = *eh;
1092 
1093 		/* XXX old crufty stuff, needs to be removed */
1094 		m_adj(m, sizeof(struct ether_header));
1095 
1096 		if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
1097 			m_freem(m);
1098 			return;
1099 		}
1100 
1101 		ether_restore_header(&m, eh, &save_eh);
1102 		if (m == NULL)
1103 			return;
1104 		eh = mtod(m, struct ether_header *);
1105 	}
1106 
1107 	ether_type = ntohs(eh->ether_type);
1108 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1109 
1110 	if (m->m_flags & M_VLANTAG) {
1111 		if (vlan_input_p != NULL) {
1112 			vlan_input_p(m);
1113 		} else {
1114 			m->m_pkthdr.rcvif->if_noproto++;
1115 			m_freem(m);
1116 		}
1117 		return;
1118 	}
1119 
1120 	m_adj(m, sizeof(struct ether_header));
1121 	redispatch = 0;
1122 
1123 	switch (ether_type) {
1124 #ifdef INET
1125 	case ETHERTYPE_IP:
1126 		if (ipflow_fastforward(m))
1127 			return;
1128 		isr = NETISR_IP;
1129 		break;
1130 
1131 	case ETHERTYPE_ARP:
1132 		if (ifp->if_flags & IFF_NOARP) {
1133 			/* Discard packet if ARP is disabled on interface */
1134 			m_freem(m);
1135 			return;
1136 		}
1137 		isr = NETISR_ARP;
1138 		break;
1139 #endif
1140 
1141 #ifdef INET6
1142 	case ETHERTYPE_IPV6:
1143 		isr = NETISR_IPV6;
1144 		break;
1145 #endif
1146 
1147 #ifdef IPX
1148 	case ETHERTYPE_IPX:
1149 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
1150 			return;
1151 		isr = NETISR_IPX;
1152 		break;
1153 #endif
1154 
1155 #ifdef NS
1156 	case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
1157 		isr = NETISR_NS;
1158 		break;
1159 
1160 #endif
1161 
1162 #ifdef NETATALK
1163 	case ETHERTYPE_AT:
1164 		isr = NETISR_ATALK1;
1165 		break;
1166 	case ETHERTYPE_AARP:
1167 		isr = NETISR_AARP;
1168 		break;
1169 #endif
1170 
1171 #ifdef MPLS
1172 	case ETHERTYPE_MPLS:
1173 	case ETHERTYPE_MPLS_MCAST:
1174 		/* Should have been set by ether_input_chain(). */
1175 		KKASSERT(m->m_flags & M_MPLSLABELED);
1176 		isr = NETISR_MPLS;
1177 		break;
1178 #endif
1179 
1180 	default:
1181 		/*
1182 		 * The accurate msgport is not determined before
1183 		 * we reach here, so redo the dispatching
1184 		 */
1185 		redispatch = 1;
1186 #ifdef IPX
1187 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
1188 			return;
1189 #endif
1190 #ifdef NS
1191 		checksum = mtod(m, ushort *);
1192 		/* Novell 802.3 */
1193 		if ((ether_type <= ETHERMTU) &&
1194 		    ((*checksum == 0xffff) || (*checksum == 0xE0E0))) {
1195 			if (*checksum == 0xE0E0) {
1196 				m->m_pkthdr.len -= 3;
1197 				m->m_len -= 3;
1198 				m->m_data += 3;
1199 			}
1200 			isr = NETISR_NS;
1201 			break;
1202 		}
1203 #endif
1204 #ifdef NETATALK
1205 		if (ether_type > ETHERMTU)
1206 			goto dropanyway;
1207 		l = mtod(m, struct llc *);
1208 		if (l->llc_dsap == LLC_SNAP_LSAP &&
1209 		    l->llc_ssap == LLC_SNAP_LSAP &&
1210 		    l->llc_control == LLC_UI) {
1211 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
1212 				 sizeof at_org_code) == 0 &&
1213 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
1214 				m_adj(m, sizeof(struct llc));
1215 				isr = NETISR_ATALK2;
1216 				break;
1217 			}
1218 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
1219 				 sizeof aarp_org_code) == 0 &&
1220 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
1221 				m_adj(m, sizeof(struct llc));
1222 				isr = NETISR_AARP;
1223 				break;
1224 			}
1225 		}
1226 dropanyway:
1227 #endif
1228 		if (ng_ether_input_orphan_p != NULL)
1229 			ng_ether_input_orphan_p(ifp, m, eh);
1230 		else
1231 			m_freem(m);
1232 		return;
1233 	}
1234 
1235 	if (!redispatch)
1236 		netisr_run(isr, m);
1237 	else
1238 		netisr_dispatch(isr, m);
1239 }
1240 
1241 /*
1242  * First we perform any link layer operations, then continue to the
1243  * upper layers with ether_demux_oncpu().
1244  */
1245 void
1246 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1247 {
1248 	if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1249 		/*
1250 		 * Receiving interface's flags are changed, when this
1251 		 * packet is waiting for processing; discard it.
1252 		 */
1253 		m_freem(m);
1254 		return;
1255 	}
1256 
1257 	/*
1258 	 * Tap the packet off here for a bridge.  bridge_input()
1259 	 * will return NULL if it has consumed the packet, otherwise
1260 	 * it gets processed as normal.  Note that bridge_input()
1261 	 * will always return the original packet if we need to
1262 	 * process it locally.
1263 	 */
1264 	if (ifp->if_bridge) {
1265 		KASSERT(bridge_input_p != NULL,
1266 			("%s: if_bridge not loaded!", __func__));
1267 
1268 		if(m->m_flags & M_PROTO1) {
1269 			m->m_flags &= ~M_PROTO1;
1270 		} else {
1271 			/* clear M_PROMISC, in case the packets comes from a vlan */
1272 			/* m->m_flags &= ~M_PROMISC; */
1273 			m = bridge_input_p(ifp, m);
1274 			if (m == NULL)
1275 				return;
1276 
1277 			KASSERT(ifp == m->m_pkthdr.rcvif,
1278 				("bridge_input_p changed rcvif\n"));
1279 		}
1280 	}
1281 
1282 	/* Handle ng_ether(4) processing, if any */
1283 	if (ng_ether_input_p != NULL) {
1284 		ng_ether_input_p(ifp, &m);
1285 		if (m == NULL)
1286 			return;
1287 	}
1288 
1289 	/* Continue with upper layer processing */
1290 	ether_demux_oncpu(ifp, m);
1291 }
1292 
1293 static void
1294 ether_input_handler(struct netmsg *nmsg)
1295 {
1296 	struct netmsg_packet *nmp = (struct netmsg_packet *)nmsg;
1297 	struct ifnet *ifp;
1298 	struct mbuf *m;
1299 
1300 	m = nmp->nm_packet;
1301 	M_ASSERTPKTHDR(m);
1302 	ifp = m->m_pkthdr.rcvif;
1303 
1304 	ether_input_oncpu(ifp, m);
1305 }
1306 
1307 static __inline void
1308 ether_init_netpacket(int num, struct mbuf *m)
1309 {
1310 	struct netmsg_packet *pmsg;
1311 
1312 	pmsg = &m->m_hdr.mh_netmsg;
1313 	netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport, 0,
1314 		    ether_input_handler);
1315 	pmsg->nm_packet = m;
1316 	pmsg->nm_netmsg.nm_lmsg.u.ms_result = num;
1317 }
1318 
1319 static __inline struct lwkt_port *
1320 ether_mport(int num, struct mbuf **m)
1321 {
1322 	if (num == NETISR_MAX) {
1323 		/*
1324 		 * All packets whose target msgports can't be
1325 		 * determined here are dispatched to netisr0,
1326 		 * where further dispatching may happen.
1327 		 */
1328 		return cpu_portfn(0);
1329 	}
1330 	return netisr_find_port(num, m);
1331 }
1332 
1333 /*
1334  * Process a received Ethernet packet.
1335  *
1336  * The ethernet header is assumed to be in the mbuf so the caller
1337  * MUST MAKE SURE that there are at least sizeof(struct ether_header)
1338  * bytes in the first mbuf.
1339  *
1340  * We first try to find the target msgport for this ether frame, if
1341  * there is no target msgport for it, this ether frame is discarded,
1342  * else we do following processing according to whether 'chain' is
1343  * NULL or not:
1344  * - If 'chain' is NULL, this ether frame is sent to the target msgport
1345  *   immediately.  This situation happens when ether_input_chain is
1346  *   accessed through ifnet.if_input.
1347  * - If 'chain' is not NULL, this ether frame is queued to the 'chain'
1348  *   bucket indexed by the target msgport's cpuid and the target msgport
1349  *   is saved in mbuf's m_pkthdr.m_head.  Caller of ether_input_chain
1350  *   must initialize 'chain' by calling ether_input_chain_init().
1351  *   ether_input_dispatch must be called later to send ether frames
1352  *   queued on 'chain' to their target msgport.
1353  */
1354 void
1355 ether_input_chain(struct ifnet *ifp, struct mbuf *m, struct mbuf_chain *chain)
1356 {
1357 	struct ether_header *eh, *save_eh, save_eh0;
1358 	struct lwkt_port *port;
1359 	uint16_t ether_type;
1360 	int isr;
1361 
1362 	ASSERT_SERIALIZED(ifp->if_serializer);
1363 	M_ASSERTPKTHDR(m);
1364 
1365 	/* Discard packet if interface is not up */
1366 	if (!(ifp->if_flags & IFF_UP)) {
1367 		m_freem(m);
1368 		return;
1369 	}
1370 
1371 	if (m->m_len < sizeof(struct ether_header)) {
1372 		/* XXX error in the caller. */
1373 		m_freem(m);
1374 		return;
1375 	}
1376 	eh = mtod(m, struct ether_header *);
1377 
1378 	m->m_pkthdr.rcvif = ifp;
1379 
1380 	logether(chain_beg, ifp);
1381 
1382 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1383 		if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1384 			 ifp->if_addrlen) == 0)
1385 			m->m_flags |= M_BCAST;
1386 		else
1387 			m->m_flags |= M_MCAST;
1388 		ifp->if_imcasts++;
1389 	}
1390 
1391 	ETHER_BPF_MTAP(ifp, m);
1392 
1393 	ifp->if_ibytes += m->m_pkthdr.len;
1394 
1395 	if (ifp->if_flags & IFF_MONITOR) {
1396 		/*
1397 		 * Interface marked for monitoring; discard packet.
1398 		 */
1399 		m_freem(m);
1400 
1401 		logether(chain_end, ifp);
1402 		return;
1403 	}
1404 
1405 	if (ntohs(eh->ether_type) == ETHERTYPE_VLAN &&
1406 	    (m->m_flags & M_VLANTAG) == 0) {
1407 		/*
1408 		 * Extract vlan tag if hardware does not do it for us
1409 		 */
1410 		vlan_ether_decap(&m);
1411 		if (m == NULL)
1412 			return;
1413 		eh = mtod(m, struct ether_header *);
1414 	}
1415 	ether_type = ntohs(eh->ether_type);
1416 
1417 	if ((m->m_flags & M_VLANTAG) && ether_type == ETHERTYPE_VLAN) {
1418 		/*
1419 		 * To prevent possible dangerous recursion,
1420 		 * we don't do vlan-in-vlan
1421 		 */
1422 		ifp->if_noproto++;
1423 		m_freem(m);
1424 		return;
1425 	}
1426 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1427 
1428 	/*
1429 	 * Map ether type to netisr id.
1430 	 */
1431 	switch (ether_type) {
1432 #ifdef INET
1433 	case ETHERTYPE_IP:
1434 		isr = NETISR_IP;
1435 		break;
1436 
1437 	case ETHERTYPE_ARP:
1438 		isr = NETISR_ARP;
1439 		break;
1440 #endif
1441 
1442 #ifdef INET6
1443 	case ETHERTYPE_IPV6:
1444 		isr = NETISR_IPV6;
1445 		break;
1446 #endif
1447 
1448 #ifdef IPX
1449 	case ETHERTYPE_IPX:
1450 		isr = NETISR_IPX;
1451 		break;
1452 #endif
1453 
1454 #ifdef NS
1455 	case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
1456 		isr = NETISR_NS;
1457 		break;
1458 #endif
1459 
1460 #ifdef NETATALK
1461 	case ETHERTYPE_AT:
1462 		isr = NETISR_ATALK1;
1463 		break;
1464 	case ETHERTYPE_AARP:
1465 		isr = NETISR_AARP;
1466 		break;
1467 #endif
1468 
1469 #ifdef MPLS
1470 	case ETHERTYPE_MPLS:
1471 	case ETHERTYPE_MPLS_MCAST:
1472 		m->m_flags |= M_MPLSLABELED;
1473 		isr = NETISR_MPLS;
1474 		break;
1475 #endif
1476 
1477 	default:
1478 		/*
1479 		 * NETISR_MAX is an invalid value; it is chosen to let
1480 		 * ether_mport() know that we are not able to decide
1481 		 * this packet's msgport here.
1482 		 */
1483 		isr = NETISR_MAX;
1484 		break;
1485 	}
1486 
1487 	/*
1488 	 * If the packet is in contiguous memory, following
1489 	 * m_adj() could ensure that the hidden ether header
1490 	 * will not be destroyed, else we will have to save
1491 	 * the ether header for the later restoration.
1492 	 */
1493 	if (m->m_pkthdr.len != m->m_len) {
1494 		save_eh0 = *eh;
1495 		save_eh = &save_eh0;
1496 	} else {
1497 		save_eh = NULL;
1498 	}
1499 
1500 	/*
1501 	 * Temporarily remove ether header; ether_mport()
1502 	 * expects a packet without ether header.
1503 	 */
1504 	m_adj(m, sizeof(struct ether_header));
1505 
1506 	/*
1507 	 * Find the packet's target msgport.
1508 	 */
1509 	port = ether_mport(isr, &m);
1510 	if (port == NULL) {
1511 		KKASSERT(m == NULL);
1512 		return;
1513 	}
1514 
1515 	/*
1516 	 * Restore ether header.
1517 	 */
1518 	if (save_eh != NULL) {
1519 		ether_restore_header(&m, eh, save_eh);
1520 		if (m == NULL)
1521 			return;
1522 	} else {
1523 		m->m_data -= ETHER_HDR_LEN;
1524 		m->m_len += ETHER_HDR_LEN;
1525 		m->m_pkthdr.len += ETHER_HDR_LEN;
1526 	}
1527 
1528 	/*
1529 	 * Initialize mbuf's netmsg packet _after_ possible
1530 	 * ether header restoration, else the initialized
1531 	 * netmsg packet may be lost during ether header
1532 	 * restoration.
1533 	 */
1534 	ether_init_netpacket(isr, m);
1535 
1536 	if (chain != NULL) {
1537 		struct mbuf_chain *c;
1538 		int cpuid;
1539 
1540 		m->m_pkthdr.header = port; /* XXX */
1541 		cpuid = port->mpu_td->td_gd->gd_cpuid;
1542 
1543 		c = &chain[cpuid];
1544 		if (c->mc_head == NULL) {
1545 			c->mc_head = c->mc_tail = m;
1546 		} else {
1547 			c->mc_tail->m_nextpkt = m;
1548 			c->mc_tail = m;
1549 		}
1550 		m->m_nextpkt = NULL;
1551 	} else {
1552 		lwkt_sendmsg(port, &m->m_hdr.mh_netmsg.nm_netmsg.nm_lmsg);
1553 	}
1554 	logether(chain_end, ifp);
1555 }
1556