xref: /dflybsd-src/sys/net/if_ethersubr.c (revision c30fd56d214a2d2bae562ca87776bc26df22beb3)
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 1.82 2008/08/05 15:11:32 nant Exp $
36  */
37 
38 #include "opt_atalk.h"
39 #include "opt_inet.h"
40 #include "opt_inet6.h"
41 #include "opt_ipx.h"
42 #include "opt_mpls.h"
43 #include "opt_netgraph.h"
44 #include "opt_carp.h"
45 #include "opt_ethernet.h"
46 
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/globaldata.h>
50 #include <sys/kernel.h>
51 #include <sys/malloc.h>
52 #include <sys/mbuf.h>
53 #include <sys/msgport.h>
54 #include <sys/socket.h>
55 #include <sys/sockio.h>
56 #include <sys/sysctl.h>
57 #include <sys/thread.h>
58 #include <sys/thread2.h>
59 
60 #include <net/if.h>
61 #include <net/netisr.h>
62 #include <net/route.h>
63 #include <net/if_llc.h>
64 #include <net/if_dl.h>
65 #include <net/if_types.h>
66 #include <net/ifq_var.h>
67 #include <net/bpf.h>
68 #include <net/ethernet.h>
69 #include <net/vlan/if_vlan_ether.h>
70 #include <net/netmsg2.h>
71 
72 #if defined(INET) || defined(INET6)
73 #include <netinet/in.h>
74 #include <netinet/in_var.h>
75 #include <netinet/if_ether.h>
76 #include <net/ipfw/ip_fw.h>
77 #include <net/dummynet/ip_dummynet.h>
78 #endif
79 #ifdef INET6
80 #include <netinet6/nd6.h>
81 #endif
82 
83 #ifdef CARP
84 #include <netinet/ip_carp.h>
85 #endif
86 
87 #ifdef IPX
88 #include <netproto/ipx/ipx.h>
89 #include <netproto/ipx/ipx_if.h>
90 int (*ef_inputp)(struct ifnet*, const struct ether_header *eh, struct mbuf *m);
91 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst,
92 		  short *tp, int *hlen);
93 #endif
94 
95 #ifdef NS
96 #include <netns/ns.h>
97 #include <netns/ns_if.h>
98 ushort ns_nettype;
99 int ether_outputdebug = 0;
100 int ether_inputdebug = 0;
101 #endif
102 
103 #ifdef NETATALK
104 #include <netproto/atalk/at.h>
105 #include <netproto/atalk/at_var.h>
106 #include <netproto/atalk/at_extern.h>
107 
108 #define	llc_snap_org_code	llc_un.type_snap.org_code
109 #define	llc_snap_ether_type	llc_un.type_snap.ether_type
110 
111 extern u_char	at_org_code[3];
112 extern u_char	aarp_org_code[3];
113 #endif /* NETATALK */
114 
115 #ifdef MPLS
116 #include <netproto/mpls/mpls.h>
117 #endif
118 
119 /* netgraph node hooks for ng_ether(4) */
120 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
121 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp,
122 		struct mbuf *m, const struct ether_header *eh);
123 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
124 void	(*ng_ether_attach_p)(struct ifnet *ifp);
125 void	(*ng_ether_detach_p)(struct ifnet *ifp);
126 
127 void	(*vlan_input2_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 /*
170  * Ethernet output routine.
171  * Encapsulate a packet of type family for the local net.
172  * Use trailer local net encapsulation if enough data in first
173  * packet leaves a multiple of 512 bytes of data in remainder.
174  * Assumes that ifp is actually pointer to arpcom structure.
175  */
176 static int
177 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
178 	     struct rtentry *rt)
179 {
180 	struct ether_header *eh, *deh;
181 	u_char *edst;
182 	int loop_copy = 0;
183 	int hlen = ETHER_HDR_LEN;	/* link layer header length */
184 	struct arpcom *ac = IFP2AC(ifp);
185 	int error;
186 
187 	ASSERT_NOT_SERIALIZED(ifp->if_serializer);
188 
189 	if (ifp->if_flags & IFF_MONITOR)
190 		gotoerr(ENETDOWN);
191 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
192 		gotoerr(ENETDOWN);
193 
194 	M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT);
195 	if (m == NULL)
196 		return (ENOBUFS);
197 	eh = mtod(m, struct ether_header *);
198 	edst = eh->ether_dhost;
199 
200 	/*
201 	 * Fill in the destination ethernet address and frame type.
202 	 */
203 	switch (dst->sa_family) {
204 #ifdef INET
205 	case AF_INET:
206 		if (!arpresolve(ifp, rt, m, dst, edst))
207 			return (0);	/* if not yet resolved */
208 #ifdef MPLS
209 		if (m->m_flags & M_MPLSLABELED)
210 			eh->ether_type = htons(ETHERTYPE_MPLS);
211 		else
212 #endif
213 			eh->ether_type = htons(ETHERTYPE_IP);
214 		break;
215 #endif
216 #ifdef INET6
217 	case AF_INET6:
218 		if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst))
219 			return (0);		/* Something bad happenned. */
220 		eh->ether_type = htons(ETHERTYPE_IPV6);
221 		break;
222 #endif
223 #ifdef IPX
224 	case AF_IPX:
225 		if (ef_outputp != NULL) {
226 			error = ef_outputp(ifp, &m, dst, &eh->ether_type,
227 					   &hlen);
228 			if (error)
229 				goto bad;
230 		} else {
231 			eh->ether_type = htons(ETHERTYPE_IPX);
232 			bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
233 			      edst, ETHER_ADDR_LEN);
234 		}
235 		break;
236 #endif
237 #ifdef NETATALK
238 	case AF_APPLETALK: {
239 		struct at_ifaddr *aa;
240 
241 		if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL) {
242 			error = 0;	/* XXX */
243 			goto bad;
244 		}
245 		/*
246 		 * In the phase 2 case, need to prepend an mbuf for
247 		 * the llc header.  Since we must preserve the value
248 		 * of m, which is passed to us by value, we m_copy()
249 		 * the first mbuf, and use it for our llc header.
250 		 */
251 		if (aa->aa_flags & AFA_PHASE2) {
252 			struct llc llc;
253 
254 			M_PREPEND(m, sizeof(struct llc), MB_DONTWAIT);
255 			eh = mtod(m, struct ether_header *);
256 			edst = eh->ether_dhost;
257 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
258 			llc.llc_control = LLC_UI;
259 			bcopy(at_org_code, llc.llc_snap_org_code,
260 			      sizeof at_org_code);
261 			llc.llc_snap_ether_type = htons(ETHERTYPE_AT);
262 			bcopy(&llc,
263 			      mtod(m, caddr_t) + sizeof(struct ether_header),
264 			      sizeof(struct llc));
265 			eh->ether_type = htons(m->m_pkthdr.len);
266 			hlen = sizeof(struct llc) + ETHER_HDR_LEN;
267 		} else {
268 			eh->ether_type = htons(ETHERTYPE_AT);
269 		}
270 		if (!aarpresolve(ac, m, (struct sockaddr_at *)dst, edst))
271 			return (0);
272 		break;
273 	  }
274 #endif
275 #ifdef NS
276 	case AF_NS:
277 		switch(ns_nettype) {
278 		default:
279 		case 0x8137:	/* Novell Ethernet_II Ethernet TYPE II */
280 			eh->ether_type = 0x8137;
281 			break;
282 		case 0x0:	/* Novell 802.3 */
283 			eh->ether_type = htons(m->m_pkthdr.len);
284 			break;
285 		case 0xe0e0:	/* Novell 802.2 and Token-Ring */
286 			M_PREPEND(m, 3, MB_DONTWAIT);
287 			eh = mtod(m, struct ether_header *);
288 			edst = eh->ether_dhost;
289 			eh->ether_type = htons(m->m_pkthdr.len);
290 			cp = mtod(m, u_char *) + sizeof(struct ether_header);
291 			*cp++ = 0xE0;
292 			*cp++ = 0xE0;
293 			*cp++ = 0x03;
294 			break;
295 		}
296 		bcopy(&(((struct sockaddr_ns *)dst)->sns_addr.x_host), edst,
297 		      ETHER_ADDR_LEN);
298 		/*
299 		 * XXX if ns_thishost is the same as the node's ethernet
300 		 * address then just the default code will catch this anyhow.
301 		 * So I'm not sure if this next clause should be here at all?
302 		 * [JRE]
303 		 */
304 		if (bcmp(edst, &ns_thishost, ETHER_ADDR_LEN) == 0) {
305 			m->m_pkthdr.rcvif = ifp;
306 			netisr_dispatch(NETISR_NS, m);
307 			return (error);
308 		}
309 		if (bcmp(edst, &ns_broadhost, ETHER_ADDR_LEN) == 0)
310 			m->m_flags |= M_BCAST;
311 		break;
312 #endif
313 	case pseudo_AF_HDRCMPLT:
314 	case AF_UNSPEC:
315 		loop_copy = -1; /* if this is for us, don't do it */
316 		deh = (struct ether_header *)dst->sa_data;
317 		memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
318 		eh->ether_type = deh->ether_type;
319 		break;
320 
321 	default:
322 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
323 		gotoerr(EAFNOSUPPORT);
324 	}
325 
326 	if (dst->sa_family == pseudo_AF_HDRCMPLT)	/* unlikely */
327 		memcpy(eh->ether_shost,
328 		       ((struct ether_header *)dst->sa_data)->ether_shost,
329 		       ETHER_ADDR_LEN);
330 	else
331 		memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
332 
333 	/*
334 	 * Bridges require special output handling.
335 	 */
336 	if (ifp->if_bridge) {
337 		KASSERT(bridge_output_p != NULL,
338 			("%s: if_bridge not loaded!", __func__));
339 		return bridge_output_p(ifp, m);
340 	}
341 
342 	/*
343 	 * If a simplex interface, and the packet is being sent to our
344 	 * Ethernet address or a broadcast address, loopback a copy.
345 	 * XXX To make a simplex device behave exactly like a duplex
346 	 * device, we should copy in the case of sending to our own
347 	 * ethernet address (thus letting the original actually appear
348 	 * on the wire). However, we don't do that here for security
349 	 * reasons and compatibility with the original behavior.
350 	 */
351 	if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
352 		int csum_flags = 0;
353 
354 		if (m->m_pkthdr.csum_flags & CSUM_IP)
355 			csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
356 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
357 			csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
358 		if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
359 			struct mbuf *n;
360 
361 			if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) {
362 				n->m_pkthdr.csum_flags |= csum_flags;
363 				if (csum_flags & CSUM_DATA_VALID)
364 					n->m_pkthdr.csum_data = 0xffff;
365 				if_simloop(ifp, n, dst->sa_family, hlen);
366 			} else
367 				ifp->if_iqdrops++;
368 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
369 				ETHER_ADDR_LEN) == 0) {
370 			m->m_pkthdr.csum_flags |= csum_flags;
371 			if (csum_flags & CSUM_DATA_VALID)
372 				m->m_pkthdr.csum_data = 0xffff;
373 			if_simloop(ifp, m, dst->sa_family, hlen);
374 			return (0);	/* XXX */
375 		}
376 	}
377 
378 #ifdef CARP
379 	if (ifp->if_carp && (error = carp_output(ifp, m, dst, NULL)))
380 		goto bad;
381 #endif
382 
383 
384 	/* Handle ng_ether(4) processing, if any */
385 	if (ng_ether_output_p != NULL) {
386 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0)
387 			goto bad;
388 		if (m == NULL)
389 			return (0);
390 	}
391 
392 	/* Continue with link-layer output */
393 	return ether_output_frame(ifp, m);
394 
395 bad:
396 	m_freem(m);
397 	return (error);
398 }
399 
400 /*
401  * Ethernet link layer output routine to send a raw frame to the device.
402  *
403  * This assumes that the 14 byte Ethernet header is present and contiguous
404  * in the first mbuf.
405  */
406 int
407 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
408 {
409 	struct ip_fw *rule = NULL;
410 	int error = 0;
411 	struct altq_pktattr pktattr;
412 	struct m_tag *mtag;
413 
414 	ASSERT_NOT_SERIALIZED(ifp->if_serializer);
415 
416 	/* Extract info from dummynet tag */
417 	mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
418 	if (mtag != NULL) {
419 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
420 
421 		m_tag_delete(m, mtag);
422 		mtag = NULL;
423 	}
424 
425 	if (ifq_is_enabled(&ifp->if_snd))
426 		altq_etherclassify(&ifp->if_snd, m, &pktattr);
427 	crit_enter();
428 	if (IPFW_LOADED && ether_ipfw != 0) {
429 		struct ether_header save_eh, *eh;
430 
431 		eh = mtod(m, struct ether_header *);
432 		save_eh = *eh;
433 		m_adj(m, ETHER_HDR_LEN);
434 		if (!ether_ipfw_chk(&m, ifp, &rule, eh)) {
435 			crit_exit();
436 			if (m != NULL) {
437 				m_freem(m);
438 				return ENOBUFS; /* pkt dropped */
439 			} else
440 				return 0;	/* consumed e.g. in a pipe */
441 		}
442 
443 		/* packet was ok, restore the ethernet header */
444 		ether_restore_header(&m, eh, &save_eh);
445 		if (m == NULL) {
446 			crit_exit();
447 			return ENOBUFS;
448 		}
449 	}
450 	crit_exit();
451 
452 	/*
453 	 * Queue message on interface, update output statistics if
454 	 * successful, and start output if interface not yet active.
455 	 */
456 	error = ifq_dispatch(ifp, m, &pktattr);
457 	return (error);
458 }
459 
460 /*
461  * ipfw processing for ethernet packets (in and out).
462  * The second parameter is NULL from ether_demux(), and ifp from
463  * ether_output_frame().
464  */
465 static boolean_t
466 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule,
467 	       const struct ether_header *eh)
468 {
469 	struct ether_header save_eh = *eh;	/* might be a ptr in m */
470 	struct ip_fw_args args;
471 	struct m_tag *mtag;
472 	int i;
473 
474 	if (*rule != NULL && fw_one_pass)
475 		return TRUE; /* dummynet packet, already partially processed */
476 
477 	/*
478 	 * I need some amount of data to be contiguous.
479 	 */
480 	i = min((*m0)->m_pkthdr.len, max_protohdr);
481 	if ((*m0)->m_len < i) {
482 		*m0 = m_pullup(*m0, i);
483 		if (*m0 == NULL)
484 			return FALSE;
485 	}
486 
487 	args.m = *m0;		/* the packet we are looking at		*/
488 	args.oif = dst;		/* destination, if any			*/
489 	if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
490 		m_tag_delete(*m0, mtag);
491 	args.rule = *rule;	/* matching rule to restart		*/
492 	args.next_hop = NULL;	/* we do not support forward yet	*/
493 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
494 	i = ip_fw_chk_ptr(&args);
495 	*m0 = args.m;
496 	*rule = args.rule;
497 
498 	if ((i & IP_FW_PORT_DENY_FLAG) || *m0 == NULL)	/* drop */
499 		return FALSE;
500 
501 	if (i == 0)					/* a PASS rule.  */
502 		return TRUE;
503 
504 	if (i & IP_FW_PORT_DYNT_FLAG) {
505 		/*
506 		 * Pass the pkt to dummynet, which consumes it.
507 		 */
508 		struct mbuf *m;
509 
510 		m = *m0;	/* pass the original to dummynet */
511 		*m0 = NULL;	/* and nothing back to the caller */
512 
513 		ether_restore_header(&m, eh, &save_eh);
514 		if (m == NULL)
515 			return FALSE;
516 
517 		ip_fw_dn_io_ptr(m, (i & 0xffff),
518 			dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
519 		return FALSE;
520 	}
521 	/*
522 	 * XXX at some point add support for divert/forward actions.
523 	 * If none of the above matches, we have to drop the pkt.
524 	 */
525 	return FALSE;
526 }
527 
528 static void
529 ether_input(struct ifnet *ifp, struct mbuf *m)
530 {
531 	ether_input_chain2(ifp, m, NULL);
532 }
533 
534 /*
535  * Perform common duties while attaching to interface list
536  */
537 void
538 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer)
539 {
540 	ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
541 			   serializer);
542 }
543 
544 void
545 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen,
546 		   lwkt_serialize_t serializer)
547 {
548 	struct sockaddr_dl *sdl;
549 
550 	ifp->if_type = IFT_ETHER;
551 	ifp->if_addrlen = ETHER_ADDR_LEN;
552 	ifp->if_hdrlen = ETHER_HDR_LEN;
553 	if_attach(ifp, serializer);
554 	ifp->if_mtu = ETHERMTU;
555 	if (ifp->if_baudrate == 0)
556 		ifp->if_baudrate = 10000000;
557 	ifp->if_output = ether_output;
558 	ifp->if_input = ether_input;
559 	ifp->if_resolvemulti = ether_resolvemulti;
560 	ifp->if_broadcastaddr = etherbroadcastaddr;
561 	sdl = IF_LLSOCKADDR(ifp);
562 	sdl->sdl_type = IFT_ETHER;
563 	sdl->sdl_alen = ifp->if_addrlen;
564 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
565 	/*
566 	 * XXX Keep the current drivers happy.
567 	 * XXX Remove once all drivers have been cleaned up
568 	 */
569 	if (lla != IFP2AC(ifp)->ac_enaddr)
570 		bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
571 	bpfattach(ifp, dlt, hdrlen);
572 	if (ng_ether_attach_p != NULL)
573 		(*ng_ether_attach_p)(ifp);
574 
575 	if_printf(ifp, "MAC address: %6D\n", lla, ":");
576 }
577 
578 /*
579  * Perform common duties while detaching an Ethernet interface
580  */
581 void
582 ether_ifdetach(struct ifnet *ifp)
583 {
584 	if_down(ifp);
585 
586 	if (ng_ether_detach_p != NULL)
587 		(*ng_ether_detach_p)(ifp);
588 	bpfdetach(ifp);
589 	if_detach(ifp);
590 }
591 
592 int
593 ether_ioctl(struct ifnet *ifp, int command, caddr_t data)
594 {
595 	struct ifaddr *ifa = (struct ifaddr *) data;
596 	struct ifreq *ifr = (struct ifreq *) data;
597 	int error = 0;
598 
599 #define IF_INIT(ifp) \
600 do { \
601 	if (((ifp)->if_flags & IFF_UP) == 0) { \
602 		(ifp)->if_flags |= IFF_UP; \
603 		(ifp)->if_init((ifp)->if_softc); \
604 	} \
605 } while (0)
606 
607 	ASSERT_SERIALIZED(ifp->if_serializer);
608 
609 	switch (command) {
610 	case SIOCSIFADDR:
611 		switch (ifa->ifa_addr->sa_family) {
612 #ifdef INET
613 		case AF_INET:
614 			IF_INIT(ifp);	/* before arpwhohas */
615 			arp_ifinit(ifp, ifa);
616 			break;
617 #endif
618 #ifdef IPX
619 		/*
620 		 * XXX - This code is probably wrong
621 		 */
622 		case AF_IPX:
623 			{
624 			struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr;
625 			struct arpcom *ac = IFP2AC(ifp);
626 
627 			if (ipx_nullhost(*ina))
628 				ina->x_host = *(union ipx_host *) ac->ac_enaddr;
629 			else
630 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
631 				      sizeof ac->ac_enaddr);
632 
633 			IF_INIT(ifp);	/* Set new address. */
634 			break;
635 			}
636 #endif
637 #ifdef NS
638 		/*
639 		 * XXX - This code is probably wrong
640 		 */
641 		case AF_NS:
642 		{
643 			struct ns_addr *ina = &(IA_SNS(ifa)->sns_addr);
644 			struct arpcom *ac = IFP2AC(ifp);
645 
646 			if (ns_nullhost(*ina))
647 				ina->x_host = *(union ns_host *)(ac->ac_enaddr);
648 			else
649 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
650 				      sizeof ac->ac_enaddr);
651 
652 			/*
653 			 * Set new address
654 			 */
655 			IF_INIT(ifp);
656 			break;
657 		}
658 #endif
659 		default:
660 			IF_INIT(ifp);
661 			break;
662 		}
663 		break;
664 
665 	case SIOCGIFADDR:
666 		bcopy(IFP2AC(ifp)->ac_enaddr,
667 		      ((struct sockaddr *)ifr->ifr_data)->sa_data,
668 		      ETHER_ADDR_LEN);
669 		break;
670 
671 	case SIOCSIFMTU:
672 		/*
673 		 * Set the interface MTU.
674 		 */
675 		if (ifr->ifr_mtu > ETHERMTU) {
676 			error = EINVAL;
677 		} else {
678 			ifp->if_mtu = ifr->ifr_mtu;
679 		}
680 		break;
681 	default:
682 		error = EINVAL;
683 		break;
684 	}
685 	return (error);
686 
687 #undef IF_INIT
688 }
689 
690 int
691 ether_resolvemulti(
692 	struct ifnet *ifp,
693 	struct sockaddr **llsa,
694 	struct sockaddr *sa)
695 {
696 	struct sockaddr_dl *sdl;
697 	struct sockaddr_in *sin;
698 #ifdef INET6
699 	struct sockaddr_in6 *sin6;
700 #endif
701 	u_char *e_addr;
702 
703 	switch(sa->sa_family) {
704 	case AF_LINK:
705 		/*
706 		 * No mapping needed. Just check that it's a valid MC address.
707 		 */
708 		sdl = (struct sockaddr_dl *)sa;
709 		e_addr = LLADDR(sdl);
710 		if ((e_addr[0] & 1) != 1)
711 			return EADDRNOTAVAIL;
712 		*llsa = 0;
713 		return 0;
714 
715 #ifdef INET
716 	case AF_INET:
717 		sin = (struct sockaddr_in *)sa;
718 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
719 			return EADDRNOTAVAIL;
720 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
721 		       M_WAITOK | M_ZERO);
722 		sdl->sdl_len = sizeof *sdl;
723 		sdl->sdl_family = AF_LINK;
724 		sdl->sdl_index = ifp->if_index;
725 		sdl->sdl_type = IFT_ETHER;
726 		sdl->sdl_alen = ETHER_ADDR_LEN;
727 		e_addr = LLADDR(sdl);
728 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
729 		*llsa = (struct sockaddr *)sdl;
730 		return 0;
731 #endif
732 #ifdef INET6
733 	case AF_INET6:
734 		sin6 = (struct sockaddr_in6 *)sa;
735 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
736 			/*
737 			 * An IP6 address of 0 means listen to all
738 			 * of the Ethernet multicast address used for IP6.
739 			 * (This is used for multicast routers.)
740 			 */
741 			ifp->if_flags |= IFF_ALLMULTI;
742 			*llsa = 0;
743 			return 0;
744 		}
745 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
746 			return EADDRNOTAVAIL;
747 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
748 		       M_WAITOK | M_ZERO);
749 		sdl->sdl_len = sizeof *sdl;
750 		sdl->sdl_family = AF_LINK;
751 		sdl->sdl_index = ifp->if_index;
752 		sdl->sdl_type = IFT_ETHER;
753 		sdl->sdl_alen = ETHER_ADDR_LEN;
754 		e_addr = LLADDR(sdl);
755 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
756 		*llsa = (struct sockaddr *)sdl;
757 		return 0;
758 #endif
759 
760 	default:
761 		/*
762 		 * Well, the text isn't quite right, but it's the name
763 		 * that counts...
764 		 */
765 		return EAFNOSUPPORT;
766 	}
767 }
768 
769 #if 0
770 /*
771  * This is for reference.  We have a table-driven version
772  * of the little-endian crc32 generator, which is faster
773  * than the double-loop.
774  */
775 uint32_t
776 ether_crc32_le(const uint8_t *buf, size_t len)
777 {
778 	uint32_t c, crc, carry;
779 	size_t i, j;
780 
781 	crc = 0xffffffffU;	/* initial value */
782 
783 	for (i = 0; i < len; i++) {
784 		c = buf[i];
785 		for (j = 0; j < 8; j++) {
786 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
787 			crc >>= 1;
788 			c >>= 1;
789 			if (carry)
790 				crc = (crc ^ ETHER_CRC_POLY_LE);
791 		}
792 	}
793 
794 	return (crc);
795 }
796 #else
797 uint32_t
798 ether_crc32_le(const uint8_t *buf, size_t len)
799 {
800 	static const uint32_t crctab[] = {
801 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
802 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
803 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
804 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
805 	};
806 	uint32_t crc;
807 	size_t i;
808 
809 	crc = 0xffffffffU;	/* initial value */
810 
811 	for (i = 0; i < len; i++) {
812 		crc ^= buf[i];
813 		crc = (crc >> 4) ^ crctab[crc & 0xf];
814 		crc = (crc >> 4) ^ crctab[crc & 0xf];
815 	}
816 
817 	return (crc);
818 }
819 #endif
820 
821 uint32_t
822 ether_crc32_be(const uint8_t *buf, size_t len)
823 {
824 	uint32_t c, crc, carry;
825 	size_t i, j;
826 
827 	crc = 0xffffffffU;	/* initial value */
828 
829 	for (i = 0; i < len; i++) {
830 		c = buf[i];
831 		for (j = 0; j < 8; j++) {
832 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
833 			crc <<= 1;
834 			c >>= 1;
835 			if (carry)
836 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
837 		}
838 	}
839 
840 	return (crc);
841 }
842 
843 /*
844  * find the size of ethernet header, and call classifier
845  */
846 void
847 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
848 		   struct altq_pktattr *pktattr)
849 {
850 	struct ether_header *eh;
851 	uint16_t ether_type;
852 	int hlen, af, hdrsize;
853 	caddr_t hdr;
854 
855 	hlen = sizeof(struct ether_header);
856 	eh = mtod(m, struct ether_header *);
857 
858 	ether_type = ntohs(eh->ether_type);
859 	if (ether_type < ETHERMTU) {
860 		/* ick! LLC/SNAP */
861 		struct llc *llc = (struct llc *)(eh + 1);
862 		hlen += 8;
863 
864 		if (m->m_len < hlen ||
865 		    llc->llc_dsap != LLC_SNAP_LSAP ||
866 		    llc->llc_ssap != LLC_SNAP_LSAP ||
867 		    llc->llc_control != LLC_UI)
868 			goto bad;  /* not snap! */
869 
870 		ether_type = ntohs(llc->llc_un.type_snap.ether_type);
871 	}
872 
873 	if (ether_type == ETHERTYPE_IP) {
874 		af = AF_INET;
875 		hdrsize = 20;  /* sizeof(struct ip) */
876 #ifdef INET6
877 	} else if (ether_type == ETHERTYPE_IPV6) {
878 		af = AF_INET6;
879 		hdrsize = 40;  /* sizeof(struct ip6_hdr) */
880 #endif
881 	} else
882 		goto bad;
883 
884 	while (m->m_len <= hlen) {
885 		hlen -= m->m_len;
886 		m = m->m_next;
887 	}
888 	hdr = m->m_data + hlen;
889 	if (m->m_len < hlen + hdrsize) {
890 		/*
891 		 * ip header is not in a single mbuf.  this should not
892 		 * happen in the current code.
893 		 * (todo: use m_pulldown in the future)
894 		 */
895 		goto bad;
896 	}
897 	m->m_data += hlen;
898 	m->m_len -= hlen;
899 	ifq_classify(ifq, m, af, pktattr);
900 	m->m_data -= hlen;
901 	m->m_len += hlen;
902 
903 	return;
904 
905 bad:
906 	pktattr->pattr_class = NULL;
907 	pktattr->pattr_hdr = NULL;
908 	pktattr->pattr_af = AF_UNSPEC;
909 }
910 
911 static void
912 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
913 		     const struct ether_header *save_eh)
914 {
915 	struct mbuf *m = *m0;
916 
917 	ether_restore_hdr++;
918 
919 	/*
920 	 * Prepend the header, optimize for the common case of
921 	 * eh pointing into the mbuf.
922 	 */
923 	if ((const void *)(eh + 1) == (void *)m->m_data) {
924 		m->m_data -= ETHER_HDR_LEN;
925 		m->m_len += ETHER_HDR_LEN;
926 		m->m_pkthdr.len += ETHER_HDR_LEN;
927 	} else {
928 		ether_prepend_hdr++;
929 
930 		M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
931 		if (m != NULL) {
932 			bcopy(save_eh, mtod(m, struct ether_header *),
933 			      ETHER_HDR_LEN);
934 		}
935 	}
936 	*m0 = m;
937 }
938 
939 #ifdef ETHER_INPUT_CHAIN
940 
941 static void
942 ether_input_ipifunc(void *arg)
943 {
944 	struct mbuf *m, *next;
945 	lwkt_port_t port;
946 
947 	m = arg;
948 	do {
949 		next = m->m_nextpkt;
950 		m->m_nextpkt = NULL;
951 
952 		port = m->m_pkthdr.header;
953 		m->m_pkthdr.header = NULL;
954 
955 		lwkt_sendmsg(port,
956 		&m->m_hdr.mh_netmsg.nm_netmsg.nm_lmsg);
957 
958 		m = next;
959 	} while (m != NULL);
960 }
961 
962 void
963 ether_input_dispatch(struct mbuf_chain *chain)
964 {
965 #ifdef SMP
966 	int i;
967 
968 	for (i = 0; i < ncpus; ++i) {
969 		if (chain[i].mc_head != NULL) {
970 			lwkt_send_ipiq(globaldata_find(i),
971 			ether_input_ipifunc, chain[i].mc_head);
972 		}
973 	}
974 #else
975 	if (chain->mc_head != NULL)
976 		ether_input_ipifunc(chain->mc_head);
977 #endif
978 }
979 
980 void
981 ether_input_chain_init(struct mbuf_chain *chain)
982 {
983 #ifdef SMP
984 	int i;
985 
986 	for (i = 0; i < ncpus; ++i)
987 		chain[i].mc_head = chain[i].mc_tail = NULL;
988 #else
989 	chain->mc_head = chain->mc_tail = NULL;
990 #endif
991 }
992 
993 #endif	/* ETHER_INPUT_CHAIN */
994 
995 /*
996  * Upper layer processing for a received Ethernet packet.
997  */
998 void
999 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
1000 {
1001 	struct ether_header *eh;
1002 	int isr, redispatch;
1003 	u_short ether_type;
1004 	struct ip_fw *rule = NULL;
1005 	struct m_tag *mtag;
1006 #ifdef NETATALK
1007 	struct llc *l;
1008 #endif
1009 
1010 	M_ASSERTPKTHDR(m);
1011 	KASSERT(m->m_len >= ETHER_HDR_LEN,
1012 		("ether header is no contiguous!\n"));
1013 
1014 	eh = mtod(m, struct ether_header *);
1015 
1016 	/* Extract info from dummynet tag */
1017 	mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
1018 	if (mtag != NULL) {
1019 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
1020 		KKASSERT(ifp == NULL);
1021 		ifp = m->m_pkthdr.rcvif;
1022 
1023 		m_tag_delete(m, mtag);
1024 		mtag = NULL;
1025 	}
1026 	if (rule)	/* packet is passing the second time */
1027 		goto post_stats;
1028 
1029 #ifdef CARP
1030 	/*
1031 	 * XXX: Okay, we need to call carp_forus() and - if it is for
1032 	 * us jump over code that does the normal check
1033 	 * "ac_enaddr == ether_dhost". The check sequence is a bit
1034 	 * different from OpenBSD, so we jump over as few code as
1035 	 * possible, to catch _all_ sanity checks. This needs
1036 	 * evaluation, to see if the carp ether_dhost values break any
1037 	 * of these checks!
1038 	 */
1039 	if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost))
1040 		goto post_stats;
1041 #endif
1042 
1043 	/*
1044 	 * Discard packet if upper layers shouldn't see it because
1045 	 * it was unicast to a different Ethernet address.  If the
1046 	 * driver is working properly, then this situation can only
1047 	 * happen when the interface is in promiscuous mode.
1048 	 */
1049 	if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
1050 	    (eh->ether_dhost[0] & 1) == 0 &&
1051 	    bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
1052 		m_freem(m);
1053 		return;
1054 	}
1055 
1056 post_stats:
1057 	if (IPFW_LOADED && ether_ipfw != 0) {
1058 		struct ether_header save_eh = *eh;
1059 
1060 		/* XXX old crufty stuff, needs to be removed */
1061 		m_adj(m, sizeof(struct ether_header));
1062 
1063 		if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
1064 			m_freem(m);
1065 			return;
1066 		}
1067 
1068 		ether_restore_header(&m, eh, &save_eh);
1069 		if (m == NULL)
1070 			return;
1071 		eh = mtod(m, struct ether_header *);
1072 	}
1073 
1074 	ether_type = ntohs(eh->ether_type);
1075 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1076 
1077 	if (m->m_flags & M_VLANTAG) {
1078 		if (vlan_input2_p != NULL) {
1079 			vlan_input2_p(m);
1080 		} else {
1081 			m->m_pkthdr.rcvif->if_noproto++;
1082 			m_freem(m);
1083 		}
1084 		return;
1085 	}
1086 
1087 	m_adj(m, sizeof(struct ether_header));
1088 	redispatch = 0;
1089 
1090 	switch (ether_type) {
1091 #ifdef INET
1092 	case ETHERTYPE_IP:
1093 		if (ipflow_fastforward(m))
1094 			return;
1095 		isr = NETISR_IP;
1096 		break;
1097 
1098 	case ETHERTYPE_ARP:
1099 		if (ifp->if_flags & IFF_NOARP) {
1100 			/* Discard packet if ARP is disabled on interface */
1101 			m_freem(m);
1102 			return;
1103 		}
1104 		isr = NETISR_ARP;
1105 		break;
1106 #endif
1107 
1108 #ifdef INET6
1109 	case ETHERTYPE_IPV6:
1110 		isr = NETISR_IPV6;
1111 		break;
1112 #endif
1113 
1114 #ifdef IPX
1115 	case ETHERTYPE_IPX:
1116 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
1117 			return;
1118 		isr = NETISR_IPX;
1119 		break;
1120 #endif
1121 
1122 #ifdef NS
1123 	case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
1124 		isr = NETISR_NS;
1125 		break;
1126 
1127 #endif
1128 
1129 #ifdef NETATALK
1130 	case ETHERTYPE_AT:
1131 		isr = NETISR_ATALK1;
1132 		break;
1133 	case ETHERTYPE_AARP:
1134 		isr = NETISR_AARP;
1135 		break;
1136 #endif
1137 
1138 #ifdef MPLS
1139 	case ETHERTYPE_MPLS:
1140 	case ETHERTYPE_MPLS_MCAST:
1141 		/* Should have been set by ether_input_chain2(). */
1142 		KKASSERT(m->m_flags & M_MPLSLABELED);
1143 		isr = NETISR_MPLS;
1144 		break;
1145 #endif
1146 
1147 	default:
1148 		/*
1149 		 * The accurate msgport is not determined before
1150 		 * we reach here, so redo the dispatching
1151 		 */
1152 		redispatch = 1;
1153 #ifdef IPX
1154 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
1155 			return;
1156 #endif
1157 #ifdef NS
1158 		checksum = mtod(m, ushort *);
1159 		/* Novell 802.3 */
1160 		if ((ether_type <= ETHERMTU) &&
1161 		    ((*checksum == 0xffff) || (*checksum == 0xE0E0))) {
1162 			if (*checksum == 0xE0E0) {
1163 				m->m_pkthdr.len -= 3;
1164 				m->m_len -= 3;
1165 				m->m_data += 3;
1166 			}
1167 			isr = NETISR_NS;
1168 			break;
1169 		}
1170 #endif
1171 #ifdef NETATALK
1172 		if (ether_type > ETHERMTU)
1173 			goto dropanyway;
1174 		l = mtod(m, struct llc *);
1175 		if (l->llc_dsap == LLC_SNAP_LSAP &&
1176 		    l->llc_ssap == LLC_SNAP_LSAP &&
1177 		    l->llc_control == LLC_UI) {
1178 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
1179 				 sizeof at_org_code) == 0 &&
1180 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
1181 				m_adj(m, sizeof(struct llc));
1182 				isr = NETISR_ATALK2;
1183 				break;
1184 			}
1185 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
1186 				 sizeof aarp_org_code) == 0 &&
1187 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
1188 				m_adj(m, sizeof(struct llc));
1189 				isr = NETISR_AARP;
1190 				break;
1191 			}
1192 		}
1193 dropanyway:
1194 #endif
1195 		if (ng_ether_input_orphan_p != NULL)
1196 			ng_ether_input_orphan_p(ifp, m, eh);
1197 		else
1198 			m_freem(m);
1199 		return;
1200 	}
1201 
1202 	if (!redispatch)
1203 		netisr_run(isr, m);
1204 	else
1205 		netisr_dispatch(isr, m);
1206 }
1207 
1208 /*
1209  * First we perform any link layer operations, then continue to the
1210  * upper layers with ether_demux_oncpu().
1211  */
1212 void
1213 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1214 {
1215 	if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1216 		/*
1217 		 * Receiving interface's flags are changed, when this
1218 		 * packet is waiting for processing; discard it.
1219 		 */
1220 		m_freem(m);
1221 		return;
1222 	}
1223 
1224 	/*
1225 	 * Tap the packet off here for a bridge.  bridge_input()
1226 	 * will return NULL if it has consumed the packet, otherwise
1227 	 * it gets processed as normal.  Note that bridge_input()
1228 	 * will always return the original packet if we need to
1229 	 * process it locally.
1230 	 */
1231 	if (ifp->if_bridge) {
1232 		KASSERT(bridge_input_p != NULL,
1233 			("%s: if_bridge not loaded!", __func__));
1234 
1235 		if(m->m_flags & M_PROTO1) {
1236 			m->m_flags &= ~M_PROTO1;
1237 		} else {
1238 			/* clear M_PROMISC, in case the packets comes from a vlan */
1239 			/* m->m_flags &= ~M_PROMISC; */
1240 			m = bridge_input_p(ifp, m);
1241 			if (m == NULL)
1242 				return;
1243 
1244 			KASSERT(ifp == m->m_pkthdr.rcvif,
1245 				("bridge_input_p changed rcvif\n"));
1246 		}
1247 	}
1248 
1249 	/* Handle ng_ether(4) processing, if any */
1250 	if (ng_ether_input_p != NULL) {
1251 		ng_ether_input_p(ifp, &m);
1252 		if (m == NULL)
1253 			return;
1254 	}
1255 
1256 	/* Continue with upper layer processing */
1257 	ether_demux_oncpu(ifp, m);
1258 }
1259 
1260 static void
1261 ether_input_handler(struct netmsg *nmsg)
1262 {
1263 	struct netmsg_packet *nmp = (struct netmsg_packet *)nmsg;
1264 	struct ifnet *ifp;
1265 	struct mbuf *m;
1266 
1267 	m = nmp->nm_packet;
1268 	M_ASSERTPKTHDR(m);
1269 	ifp = m->m_pkthdr.rcvif;
1270 
1271 	ether_input_oncpu(ifp, m);
1272 }
1273 
1274 static __inline void
1275 ether_init_netpacket(int num, struct mbuf *m)
1276 {
1277 	struct netmsg_packet *pmsg;
1278 
1279 	pmsg = &m->m_hdr.mh_netmsg;
1280 	netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport, 0,
1281 		    ether_input_handler);
1282 	pmsg->nm_packet = m;
1283 	pmsg->nm_netmsg.nm_lmsg.u.ms_result = num;
1284 }
1285 
1286 static __inline struct lwkt_port *
1287 ether_mport(int num, struct mbuf **m)
1288 {
1289 	if (num == NETISR_MAX) {
1290 		/*
1291 		 * All packets whose target msgports can't be
1292 		 * determined here are dispatched to netisr0,
1293 		 * where further dispatching may happen.
1294 		 */
1295 		return cpu_portfn(0);
1296 	}
1297 	return netisr_find_port(num, m);
1298 }
1299 
1300 /*
1301  * Process a received Ethernet packet.
1302  *
1303  * The ethernet header is assumed to be in the mbuf so the caller
1304  * MUST MAKE SURE that there are at least sizeof(struct ether_header)
1305  * bytes in the first mbuf.
1306  *
1307  * We first try to find the target msgport for this ether frame, if
1308  * there is no target msgport for it, this ether frame is discarded,
1309  * else we do following processing according to whether 'chain' is
1310  * NULL or not:
1311  * - If 'chain' is NULL, this ether frame is sent to the target msgport
1312  *   immediately.  This situation happens when ether_input_chain2 is
1313  *   accessed through ifnet.if_input.
1314  * - If 'chain' is not NULL, this ether frame is queued to the 'chain'
1315  *   bucket indexed by the target msgport's cpuid and the target msgport
1316  *   is saved in mbuf's m_pkthdr.m_head.  Caller of ether_input_chain2
1317  *   must initialize 'chain' by calling ether_input_chain_init().
1318  *   ether_input_dispatch must be called later to send ether frames
1319  *   queued on 'chain' to their target msgport.
1320  */
1321 void
1322 ether_input_chain2(struct ifnet *ifp, struct mbuf *m, struct mbuf_chain *chain)
1323 {
1324 	struct ether_header *eh, *save_eh, save_eh0;
1325 	struct lwkt_port *port;
1326 	uint16_t ether_type;
1327 	int isr;
1328 
1329 	ASSERT_SERIALIZED(ifp->if_serializer);
1330 	M_ASSERTPKTHDR(m);
1331 
1332 	/* Discard packet if interface is not up */
1333 	if (!(ifp->if_flags & IFF_UP)) {
1334 		m_freem(m);
1335 		return;
1336 	}
1337 
1338 	if (m->m_len < sizeof(struct ether_header)) {
1339 		/* XXX error in the caller. */
1340 		m_freem(m);
1341 		return;
1342 	}
1343 	eh = mtod(m, struct ether_header *);
1344 
1345 	m->m_pkthdr.rcvif = ifp;
1346 
1347 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1348 		if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1349 			 ifp->if_addrlen) == 0)
1350 			m->m_flags |= M_BCAST;
1351 		else
1352 			m->m_flags |= M_MCAST;
1353 		ifp->if_imcasts++;
1354 	}
1355 
1356 	ETHER_BPF_MTAP(ifp, m);
1357 
1358 	ifp->if_ibytes += m->m_pkthdr.len;
1359 
1360 	if (ifp->if_flags & IFF_MONITOR) {
1361 		/*
1362 		 * Interface marked for monitoring; discard packet.
1363 		 */
1364 		m_freem(m);
1365 		return;
1366 	}
1367 
1368 	if (ntohs(eh->ether_type) == ETHERTYPE_VLAN &&
1369 	    (m->m_flags & M_VLANTAG) == 0) {
1370 		/*
1371 		 * Extract vlan tag if hardware does not do it for us
1372 		 */
1373 		vlan_ether_decap(&m);
1374 		if (m == NULL)
1375 			return;
1376 		eh = mtod(m, struct ether_header *);
1377 	}
1378 	ether_type = ntohs(eh->ether_type);
1379 
1380 	if ((m->m_flags & M_VLANTAG) && ether_type == ETHERTYPE_VLAN) {
1381 		/*
1382 		 * To prevent possible dangerous recursion,
1383 		 * we don't do vlan-in-vlan
1384 		 */
1385 		ifp->if_noproto++;
1386 		m_freem(m);
1387 		return;
1388 	}
1389 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1390 
1391 	/*
1392 	 * Map ether type to netisr id.
1393 	 */
1394 	switch (ether_type) {
1395 #ifdef INET
1396 	case ETHERTYPE_IP:
1397 		isr = NETISR_IP;
1398 		break;
1399 
1400 	case ETHERTYPE_ARP:
1401 		isr = NETISR_ARP;
1402 		break;
1403 #endif
1404 
1405 #ifdef INET6
1406 	case ETHERTYPE_IPV6:
1407 		isr = NETISR_IPV6;
1408 		break;
1409 #endif
1410 
1411 #ifdef IPX
1412 	case ETHERTYPE_IPX:
1413 		isr = NETISR_IPX;
1414 		break;
1415 #endif
1416 
1417 #ifdef NS
1418 	case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
1419 		isr = NETISR_NS;
1420 		break;
1421 #endif
1422 
1423 #ifdef NETATALK
1424 	case ETHERTYPE_AT:
1425 		isr = NETISR_ATALK1;
1426 		break;
1427 	case ETHERTYPE_AARP:
1428 		isr = NETISR_AARP;
1429 		break;
1430 #endif
1431 
1432 #ifdef MPLS
1433 	case ETHERTYPE_MPLS:
1434 	case ETHERTYPE_MPLS_MCAST:
1435 		m->m_flags |= M_MPLSLABELED;
1436 		isr = NETISR_MPLS;
1437 		break;
1438 #endif
1439 
1440 	default:
1441 		/*
1442 		 * NETISR_MAX is an invalid value; it is chosen to let
1443 		 * ether_mport() know that we are not able to decide
1444 		 * this packet's msgport here.
1445 		 */
1446 		isr = NETISR_MAX;
1447 		break;
1448 	}
1449 
1450 	/*
1451 	 * If the packet is in contiguous memory, following
1452 	 * m_adj() could ensure that the hidden ether header
1453 	 * will not be destroyed, else we will have to save
1454 	 * the ether header for the later restoration.
1455 	 */
1456 	if (m->m_pkthdr.len != m->m_len) {
1457 		save_eh0 = *eh;
1458 		save_eh = &save_eh0;
1459 	} else {
1460 		save_eh = NULL;
1461 	}
1462 
1463 	/*
1464 	 * Temporarily remove ether header; ether_mport()
1465 	 * expects a packet without ether header.
1466 	 */
1467 	m_adj(m, sizeof(struct ether_header));
1468 
1469 	/*
1470 	 * Find the packet's target msgport.
1471 	 */
1472 	port = ether_mport(isr, &m);
1473 	if (port == NULL) {
1474 		KKASSERT(m == NULL);
1475 		return;
1476 	}
1477 
1478 	/*
1479 	 * Restore ether header.
1480 	 */
1481 	if (save_eh != NULL) {
1482 		ether_restore_header(&m, eh, save_eh);
1483 		if (m == NULL)
1484 			return;
1485 	} else {
1486 		m->m_data -= ETHER_HDR_LEN;
1487 		m->m_len += ETHER_HDR_LEN;
1488 		m->m_pkthdr.len += ETHER_HDR_LEN;
1489 	}
1490 
1491 	/*
1492 	 * Initialize mbuf's netmsg packet _after_ possible
1493 	 * ether header restoration, else the initialized
1494 	 * netmsg packet may be lost during ether header
1495 	 * restoration.
1496 	 */
1497 	ether_init_netpacket(isr, m);
1498 
1499 #ifdef ETHER_INPUT_CHAIN
1500 	if (chain != NULL) {
1501 		struct mbuf_chain *c;
1502 		int cpuid;
1503 
1504 		m->m_pkthdr.header = port; /* XXX */
1505 		cpuid = port->mpu_td->td_gd->gd_cpuid;
1506 
1507 		c = &chain[cpuid];
1508 		if (c->mc_head == NULL) {
1509 			c->mc_head = c->mc_tail = m;
1510 		} else {
1511 			c->mc_tail->m_nextpkt = m;
1512 			c->mc_tail = m;
1513 		}
1514 		m->m_nextpkt = NULL;
1515 	} else
1516 #endif	/* ETHER_INPUT_CHAIN */
1517 		lwkt_sendmsg(port, &m->m_hdr.mh_netmsg.nm_netmsg.nm_lmsg);
1518 }
1519