xref: /freebsd/sys/net/if_ethersubr.c (revision 588ff6c0cc9aaf10ba19080d9f8acbd8be36abf3)
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  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)if_ethersubr.c	8.1 (Berkeley) 6/10/93
30  * $FreeBSD$
31  */
32 
33 #include "opt_atalk.h"
34 #include "opt_inet.h"
35 #include "opt_inet6.h"
36 #include "opt_ipx.h"
37 #include "opt_mac.h"
38 #include "opt_netgraph.h"
39 #include "opt_carp.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kernel.h>
44 #include <sys/malloc.h>
45 #include <sys/module.h>
46 #include <sys/mbuf.h>
47 #include <sys/random.h>
48 #include <sys/socket.h>
49 #include <sys/sockio.h>
50 #include <sys/sysctl.h>
51 
52 #include <net/if.h>
53 #include <net/if_arp.h>
54 #include <net/netisr.h>
55 #include <net/route.h>
56 #include <net/if_llc.h>
57 #include <net/if_dl.h>
58 #include <net/if_types.h>
59 #include <net/bpf.h>
60 #include <net/ethernet.h>
61 #include <net/if_bridgevar.h>
62 #include <net/if_vlan_var.h>
63 
64 #if defined(INET) || defined(INET6)
65 #include <netinet/in.h>
66 #include <netinet/in_var.h>
67 #include <netinet/if_ether.h>
68 #include <netinet/ip_fw.h>
69 #include <netinet/ip_dummynet.h>
70 #endif
71 #ifdef INET6
72 #include <netinet6/nd6.h>
73 #endif
74 
75 #ifdef DEV_CARP
76 #include <netinet/ip_carp.h>
77 #endif
78 
79 #ifdef IPX
80 #include <netipx/ipx.h>
81 #include <netipx/ipx_if.h>
82 #endif
83 int (*ef_inputp)(struct ifnet*, struct ether_header *eh, struct mbuf *m);
84 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp,
85 		struct sockaddr *dst, short *tp, int *hlen);
86 
87 #ifdef NETATALK
88 #include <netatalk/at.h>
89 #include <netatalk/at_var.h>
90 #include <netatalk/at_extern.h>
91 
92 #define llc_snap_org_code llc_un.type_snap.org_code
93 #define llc_snap_ether_type llc_un.type_snap.ether_type
94 
95 extern u_char	at_org_code[3];
96 extern u_char	aarp_org_code[3];
97 #endif /* NETATALK */
98 
99 #include <security/mac/mac_framework.h>
100 
101 /* netgraph node hooks for ng_ether(4) */
102 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
103 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
104 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
105 void	(*ng_ether_attach_p)(struct ifnet *ifp);
106 void	(*ng_ether_detach_p)(struct ifnet *ifp);
107 
108 void	(*vlan_input_p)(struct ifnet *, struct mbuf *);
109 
110 /* if_bridge(4) support */
111 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
112 int	(*bridge_output_p)(struct ifnet *, struct mbuf *,
113 		struct sockaddr *, struct rtentry *);
114 void	(*bridge_dn_p)(struct mbuf *, struct ifnet *);
115 
116 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
117 			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
118 
119 static	int ether_resolvemulti(struct ifnet *, struct sockaddr **,
120 		struct sockaddr *);
121 
122 /* XXX: should be in an arp support file, not here */
123 MALLOC_DEFINE(M_ARPCOM, "arpcom", "802.* interface internals");
124 
125 #define senderr(e) do { error = (e); goto bad;} while (0)
126 
127 #if defined(INET) || defined(INET6)
128 int
129 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
130 	struct ip_fw **rule, int shared);
131 static int ether_ipfw;
132 #endif
133 
134 /*
135  * Ethernet output routine.
136  * Encapsulate a packet of type family for the local net.
137  * Use trailer local net encapsulation if enough data in first
138  * packet leaves a multiple of 512 bytes of data in remainder.
139  */
140 int
141 ether_output(struct ifnet *ifp, struct mbuf *m,
142 	struct sockaddr *dst, struct rtentry *rt0)
143 {
144 	short type;
145 	int error, hdrcmplt = 0;
146 	u_char esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
147 	struct ether_header *eh;
148 	int loop_copy = 1;
149 	int hlen;	/* link layer header length */
150 
151 #ifdef MAC
152 	error = mac_check_ifnet_transmit(ifp, m);
153 	if (error)
154 		senderr(error);
155 #endif
156 
157 	if (ifp->if_flags & IFF_MONITOR)
158 		senderr(ENETDOWN);
159 	if (!((ifp->if_flags & IFF_UP) &&
160 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
161 		senderr(ENETDOWN);
162 
163 	hlen = ETHER_HDR_LEN;
164 	switch (dst->sa_family) {
165 #ifdef INET
166 	case AF_INET:
167 		error = arpresolve(ifp, rt0, m, dst, edst);
168 		if (error)
169 			return (error == EWOULDBLOCK ? 0 : error);
170 		type = htons(ETHERTYPE_IP);
171 		break;
172 	case AF_ARP:
173 	{
174 		struct arphdr *ah;
175 		ah = mtod(m, struct arphdr *);
176 		ah->ar_hrd = htons(ARPHRD_ETHER);
177 
178 		loop_copy = 0; /* if this is for us, don't do it */
179 
180 		switch(ntohs(ah->ar_op)) {
181 		case ARPOP_REVREQUEST:
182 		case ARPOP_REVREPLY:
183 			type = htons(ETHERTYPE_REVARP);
184 			break;
185 		case ARPOP_REQUEST:
186 		case ARPOP_REPLY:
187 		default:
188 			type = htons(ETHERTYPE_ARP);
189 			break;
190 		}
191 
192 		if (m->m_flags & M_BCAST)
193 			bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN);
194 		else
195 			bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN);
196 
197 	}
198 	break;
199 #endif
200 #ifdef INET6
201 	case AF_INET6:
202 		error = nd6_storelladdr(ifp, rt0, m, dst, (u_char *)edst);
203 		if (error)
204 			return error;
205 		type = htons(ETHERTYPE_IPV6);
206 		break;
207 #endif
208 #ifdef IPX
209 	case AF_IPX:
210 		if (ef_outputp) {
211 		    error = ef_outputp(ifp, &m, dst, &type, &hlen);
212 		    if (error)
213 			goto bad;
214 		} else
215 		    type = htons(ETHERTYPE_IPX);
216 		bcopy((caddr_t)&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
217 		    (caddr_t)edst, sizeof (edst));
218 		break;
219 #endif
220 #ifdef NETATALK
221 	case AF_APPLETALK:
222 	  {
223 	    struct at_ifaddr *aa;
224 
225 	    if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL)
226 		    senderr(EHOSTUNREACH); /* XXX */
227 	    if (!aarpresolve(ifp, m, (struct sockaddr_at *)dst, edst))
228 		    return (0);
229 	    /*
230 	     * In the phase 2 case, need to prepend an mbuf for the llc header.
231 	     */
232 	    if ( aa->aa_flags & AFA_PHASE2 ) {
233 		struct llc llc;
234 
235 		M_PREPEND(m, LLC_SNAPFRAMELEN, M_DONTWAIT);
236 		if (m == NULL)
237 			senderr(ENOBUFS);
238 		llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
239 		llc.llc_control = LLC_UI;
240 		bcopy(at_org_code, llc.llc_snap_org_code, sizeof(at_org_code));
241 		llc.llc_snap_ether_type = htons( ETHERTYPE_AT );
242 		bcopy(&llc, mtod(m, caddr_t), LLC_SNAPFRAMELEN);
243 		type = htons(m->m_pkthdr.len);
244 		hlen = LLC_SNAPFRAMELEN + ETHER_HDR_LEN;
245 	    } else {
246 		type = htons(ETHERTYPE_AT);
247 	    }
248 	    break;
249 	  }
250 #endif /* NETATALK */
251 
252 	case pseudo_AF_HDRCMPLT:
253 		hdrcmplt = 1;
254 		eh = (struct ether_header *)dst->sa_data;
255 		(void)memcpy(esrc, eh->ether_shost, sizeof (esrc));
256 		/* FALLTHROUGH */
257 
258 	case AF_UNSPEC:
259 		loop_copy = 0; /* if this is for us, don't do it */
260 		eh = (struct ether_header *)dst->sa_data;
261 		(void)memcpy(edst, eh->ether_dhost, sizeof (edst));
262 		type = eh->ether_type;
263 		break;
264 
265 	default:
266 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
267 		senderr(EAFNOSUPPORT);
268 	}
269 
270 	/*
271 	 * Add local net header.  If no space in first mbuf,
272 	 * allocate another.
273 	 */
274 	M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
275 	if (m == NULL)
276 		senderr(ENOBUFS);
277 	eh = mtod(m, struct ether_header *);
278 	(void)memcpy(&eh->ether_type, &type,
279 		sizeof(eh->ether_type));
280 	(void)memcpy(eh->ether_dhost, edst, sizeof (edst));
281 	if (hdrcmplt)
282 		(void)memcpy(eh->ether_shost, esrc,
283 			sizeof(eh->ether_shost));
284 	else
285 		(void)memcpy(eh->ether_shost, IF_LLADDR(ifp),
286 			sizeof(eh->ether_shost));
287 
288 	/*
289 	 * If a simplex interface, and the packet is being sent to our
290 	 * Ethernet address or a broadcast address, loopback a copy.
291 	 * XXX To make a simplex device behave exactly like a duplex
292 	 * device, we should copy in the case of sending to our own
293 	 * ethernet address (thus letting the original actually appear
294 	 * on the wire). However, we don't do that here for security
295 	 * reasons and compatibility with the original behavior.
296 	 */
297 	if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy &&
298 	    m_tag_find(m, PACKET_TAG_PF_ROUTED, NULL) == NULL) {
299 		int csum_flags = 0;
300 
301 		if (m->m_pkthdr.csum_flags & CSUM_IP)
302 			csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
303 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
304 			csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
305 
306 		if (m->m_flags & M_BCAST) {
307 			struct mbuf *n;
308 
309 			/*
310 			 * Because if_simloop() modifies the packet, we need a
311 			 * writable copy through m_dup() instead of a readonly
312 			 * one as m_copy[m] would give us. The alternative would
313 			 * be to modify if_simloop() to handle the readonly mbuf,
314 			 * but performancewise it is mostly equivalent (trading
315 			 * extra data copying vs. extra locking).
316 			 *
317 			 * XXX This is a local workaround.  A number of less
318 			 * often used kernel parts suffer from the same bug.
319 			 * See PR kern/105943 for a proposed general solution.
320 			 */
321 			if ((n = m_dup(m, M_DONTWAIT)) != NULL) {
322 				n->m_pkthdr.csum_flags |= csum_flags;
323 				if (csum_flags & CSUM_DATA_VALID)
324 					n->m_pkthdr.csum_data = 0xffff;
325 				(void)if_simloop(ifp, n, dst->sa_family, hlen);
326 			} else
327 				ifp->if_iqdrops++;
328 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
329 				ETHER_ADDR_LEN) == 0) {
330 			m->m_pkthdr.csum_flags |= csum_flags;
331 			if (csum_flags & CSUM_DATA_VALID)
332 				m->m_pkthdr.csum_data = 0xffff;
333 			(void) if_simloop(ifp, m, dst->sa_family, hlen);
334 			return (0);	/* XXX */
335 		}
336 	}
337 
338        /*
339 	* Bridges require special output handling.
340 	*/
341 	if (ifp->if_bridge) {
342 		BRIDGE_OUTPUT(ifp, m, error);
343 		return (error);
344 	}
345 
346 #ifdef DEV_CARP
347 	if (ifp->if_carp &&
348 	    (error = carp_output(ifp, m, dst, NULL)))
349 		goto bad;
350 #endif
351 
352 	/* Handle ng_ether(4) processing, if any */
353 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
354 		KASSERT(ng_ether_output_p != NULL,
355 		    ("ng_ether_output_p is NULL"));
356 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
357 bad:			if (m != NULL)
358 				m_freem(m);
359 			return (error);
360 		}
361 		if (m == NULL)
362 			return (0);
363 	}
364 
365 	/* Continue with link-layer output */
366 	return ether_output_frame(ifp, m);
367 }
368 
369 /*
370  * Ethernet link layer output routine to send a raw frame to the device.
371  *
372  * This assumes that the 14 byte Ethernet header is present and contiguous
373  * in the first mbuf (if BRIDGE'ing).
374  */
375 int
376 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
377 {
378 	int error;
379 #if defined(INET) || defined(INET6)
380 	struct ip_fw *rule = ip_dn_claim_rule(m);
381 
382 	if (IPFW_LOADED && ether_ipfw != 0) {
383 		if (ether_ipfw_chk(&m, ifp, &rule, 0) == 0) {
384 			if (m) {
385 				m_freem(m);
386 				return EACCES;	/* pkt dropped */
387 			} else
388 				return 0;	/* consumed e.g. in a pipe */
389 		}
390 	}
391 #endif
392 
393 	/*
394 	 * Queue message on interface, update output statistics if
395 	 * successful, and start output if interface not yet active.
396 	 */
397 	IFQ_HANDOFF(ifp, m, error);
398 	return (error);
399 }
400 
401 #if defined(INET) || defined(INET6)
402 /*
403  * ipfw processing for ethernet packets (in and out).
404  * The second parameter is NULL from ether_demux, and ifp from
405  * ether_output_frame.
406  */
407 int
408 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
409 	struct ip_fw **rule, int shared)
410 {
411 	struct ether_header *eh;
412 	struct ether_header save_eh;
413 	struct mbuf *m;
414 	int i;
415 	struct ip_fw_args args;
416 
417 	if (*rule != NULL && fw_one_pass)
418 		return 1; /* dummynet packet, already partially processed */
419 
420 	/*
421 	 * I need some amt of data to be contiguous, and in case others need
422 	 * the packet (shared==1) also better be in the first mbuf.
423 	 */
424 	m = *m0;
425 	i = min( m->m_pkthdr.len, max_protohdr);
426 	if ( shared || m->m_len < i) {
427 		m = m_pullup(m, i);
428 		if (m == NULL) {
429 			*m0 = m;
430 			return 0;
431 		}
432 	}
433 	eh = mtod(m, struct ether_header *);
434 	save_eh = *eh;			/* save copy for restore below */
435 	m_adj(m, ETHER_HDR_LEN);	/* strip ethernet header */
436 
437 	args.m = m;		/* the packet we are looking at		*/
438 	args.oif = dst;		/* destination, if any			*/
439 	args.rule = *rule;	/* matching rule to restart		*/
440 	args.next_hop = NULL;	/* we do not support forward yet	*/
441 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
442 	args.inp = NULL;	/* used by ipfw uid/gid/jail rules	*/
443 	i = ip_fw_chk_ptr(&args);
444 	m = args.m;
445 	if (m != NULL) {
446 		/*
447 		 * Restore Ethernet header, as needed, in case the
448 		 * mbuf chain was replaced by ipfw.
449 		 */
450 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
451 		if (m == NULL) {
452 			*m0 = m;
453 			return 0;
454 		}
455 		if (eh != mtod(m, struct ether_header *))
456 			bcopy(&save_eh, mtod(m, struct ether_header *),
457 				ETHER_HDR_LEN);
458 	}
459 	*m0 = m;
460 	*rule = args.rule;
461 
462 	if (i == IP_FW_DENY) /* drop */
463 		return 0;
464 
465 	KASSERT(m != NULL, ("ether_ipfw_chk: m is NULL"));
466 
467 	if (i == IP_FW_PASS) /* a PASS rule.  */
468 		return 1;
469 
470 	if (DUMMYNET_LOADED && (i == IP_FW_DUMMYNET)) {
471 		/*
472 		 * Pass the pkt to dummynet, which consumes it.
473 		 * If shared, make a copy and keep the original.
474 		 */
475 		if (shared) {
476 			m = m_copypacket(m, M_DONTWAIT);
477 			if (m == NULL)
478 				return 0;
479 		} else {
480 			/*
481 			 * Pass the original to dummynet and
482 			 * nothing back to the caller
483 			 */
484 			*m0 = NULL ;
485 		}
486 		ip_dn_io_ptr(m, dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
487 		return 0;
488 	}
489 	/*
490 	 * XXX at some point add support for divert/forward actions.
491 	 * If none of the above matches, we have to drop the pkt.
492 	 */
493 	return 0;
494 }
495 #endif
496 
497 /*
498  * Process a received Ethernet packet; the packet is in the
499  * mbuf chain m with the ethernet header at the front.
500  */
501 static void
502 ether_input(struct ifnet *ifp, struct mbuf *m)
503 {
504 	struct ether_header *eh;
505 	u_short etype;
506 
507 	/*
508 	 * Do consistency checks to verify assumptions
509 	 * made by code past this point.
510 	 */
511 	if ((m->m_flags & M_PKTHDR) == 0) {
512 		if_printf(ifp, "discard frame w/o packet header\n");
513 		ifp->if_ierrors++;
514 		m_freem(m);
515 		return;
516 	}
517 	if (m->m_len < ETHER_HDR_LEN) {
518 		/* XXX maybe should pullup? */
519 		if_printf(ifp, "discard frame w/o leading ethernet "
520 				"header (len %u pkt len %u)\n",
521 				m->m_len, m->m_pkthdr.len);
522 		ifp->if_ierrors++;
523 		m_freem(m);
524 		return;
525 	}
526 	eh = mtod(m, struct ether_header *);
527 	etype = ntohs(eh->ether_type);
528 	if (m->m_pkthdr.len >
529 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
530 		if_printf(ifp, "discard oversize frame "
531 				"(ether type %x flags %x len %u > max %lu)\n",
532 				etype, m->m_flags, m->m_pkthdr.len,
533 				ETHER_MAX_FRAME(ifp, etype,
534 						m->m_flags & M_HASFCS));
535 		ifp->if_ierrors++;
536 		m_freem(m);
537 		return;
538 	}
539 	if (m->m_pkthdr.rcvif == NULL) {
540 		if_printf(ifp, "discard frame w/o interface pointer\n");
541 		ifp->if_ierrors++;
542 		m_freem(m);
543 		return;
544 	}
545 #ifdef DIAGNOSTIC
546 	if (m->m_pkthdr.rcvif != ifp) {
547 		if_printf(ifp, "Warning, frame marked as received on %s\n",
548 			m->m_pkthdr.rcvif->if_xname);
549 	}
550 #endif
551 
552 #ifdef MAC
553 	/*
554 	 * Tag the mbuf with an appropriate MAC label before any other
555 	 * consumers can get to it.
556 	 */
557 	mac_create_mbuf_from_ifnet(ifp, m);
558 #endif
559 
560 	/*
561 	 * Give bpf a chance at the packet.  Process any 802.1Q tags
562 	 * that may have been attached to the mbuf.
563 	 */
564 	ETHER_BPF_MTAP(ifp, m);
565 
566 	/* If the CRC is still on the packet, trim it off. */
567 	if (m->m_flags & M_HASFCS) {
568 		m_adj(m, -ETHER_CRC_LEN);
569 		m->m_flags &= ~M_HASFCS;
570 	}
571 
572 	ifp->if_ibytes += m->m_pkthdr.len;
573 
574 	if (ifp->if_flags & IFF_MONITOR) {
575 		/*
576 		 * Interface marked for monitoring; discard packet.
577 		 */
578 		m_freem(m);
579 		return;
580 	}
581 
582 	/* Handle ng_ether(4) processing, if any */
583 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
584 		KASSERT(ng_ether_input_p != NULL,
585 		    ("ng_ether_input_p is NULL"));
586 		(*ng_ether_input_p)(ifp, &m);
587 		if (m == NULL)
588 			return;
589 	}
590 
591 	/*
592 	 * Tap the packet off here for a bridge.  bridge_input()
593 	 * will return NULL if it has consumed the packet, otherwise
594 	 * it gets processed as normal.  Note that bridge_input()
595 	 * will always return the original packet if we need to
596 	 * process it locally.
597 	 *
598 	 * XXX: When changing the below block, please also look
599 	 * at the src/sys/netgraph/ng_ether.c:ng_ether_rcv_upper()
600 	 */
601 	if (ifp->if_bridge) {
602 		BRIDGE_INPUT(ifp, m);
603 		if (m == NULL)
604 			return;
605 	}
606 
607 	/* First chunk of an mbuf contains good entropy */
608 	if (harvest.ethernet)
609 		random_harvest(m, 16, 3, 0, RANDOM_NET);
610 	ether_demux(ifp, m);
611 }
612 
613 /*
614  * Upper layer processing for a received Ethernet packet.
615  */
616 void
617 ether_demux(struct ifnet *ifp, struct mbuf *m)
618 {
619 	struct ether_header *eh;
620 	int isr;
621 	u_short ether_type;
622 #if defined(NETATALK)
623 	struct llc *l;
624 #endif
625 #if defined(INET) || defined(INET6)
626 	struct ip_fw *rule = ip_dn_claim_rule(m);
627 #endif
628 
629 	KASSERT(ifp != NULL, ("ether_demux: NULL interface pointer"));
630 
631 	eh = mtod(m, struct ether_header *);
632 	ether_type = ntohs(eh->ether_type);
633 
634 #if defined(INET) || defined(INET6)
635 	if (rule)	/* packet was already bridged */
636 		goto post_stats;
637 #endif
638 
639 	if (!(ifp->if_bridge) &&
640 	    !((ether_type == ETHERTYPE_VLAN || m->m_flags & M_VLANTAG) &&
641 	    ifp->if_vlantrunk != NULL)) {
642 #ifdef DEV_CARP
643 		/*
644 		 * XXX: Okay, we need to call carp_forus() and - if it is for
645 		 * us jump over code that does the normal check
646 		 * "IF_LLADDR(ifp) == ether_dhost". The check sequence is a bit
647 		 * different from OpenBSD, so we jump over as few code as
648 		 * possible, to catch _all_ sanity checks. This needs
649 		 * evaluation, to see if the carp ether_dhost values break any
650 		 * of these checks!
651 		 */
652 		if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost))
653 			goto pre_stats;
654 #endif
655 		/*
656 		 * Discard packet if upper layers shouldn't see it because it
657 		 * was unicast to a different Ethernet address. If the driver
658 		 * is working properly, then this situation can only happen
659 		 * when the interface is in promiscuous mode.
660 		 *
661 		 * If VLANs are active, and this packet has a VLAN tag, do
662 		 * not drop it here but pass it on to the VLAN layer, to
663 		 * give them a chance to consider it as well (e. g. in case
664 		 * bridging is only active on a VLAN).  They will drop it if
665 		 * it's undesired.
666 		 */
667 		if ((ifp->if_flags & IFF_PROMISC) != 0
668 		    && !ETHER_IS_MULTICAST(eh->ether_dhost)
669 		    && bcmp(eh->ether_dhost,
670 		      IF_LLADDR(ifp), ETHER_ADDR_LEN) != 0
671 		    && (ifp->if_flags & IFF_PPROMISC) == 0) {
672 			    m_freem(m);
673 			    return;
674 		}
675 	}
676 
677 #ifdef DEV_CARP
678 pre_stats:
679 #endif
680 	/* Discard packet if interface is not up */
681 	if ((ifp->if_flags & IFF_UP) == 0) {
682 		m_freem(m);
683 		return;
684 	}
685 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
686 		if (bcmp(etherbroadcastaddr, eh->ether_dhost,
687 		    sizeof(etherbroadcastaddr)) == 0)
688 			m->m_flags |= M_BCAST;
689 		else
690 			m->m_flags |= M_MCAST;
691 	}
692 	if (m->m_flags & (M_BCAST|M_MCAST))
693 		ifp->if_imcasts++;
694 
695 #if defined(INET) || defined(INET6)
696 post_stats:
697 	if (IPFW_LOADED && ether_ipfw != 0) {
698 		if (ether_ipfw_chk(&m, NULL, &rule, 0) == 0) {
699 			if (m)
700 				m_freem(m);
701 			return;
702 		}
703 	}
704 #endif
705 
706 	/*
707 	 * Check to see if the device performed the VLAN decapsulation and
708 	 * provided us with the tag.
709 	 */
710 	if (m->m_flags & M_VLANTAG) {
711 		/*
712 		 * If no VLANs are configured, drop.
713 		 */
714 		if (ifp->if_vlantrunk == NULL) {
715 			ifp->if_noproto++;
716 			m_freem(m);
717 			return;
718 		}
719 		/*
720 		 * vlan_input() will either recursively call ether_input()
721 		 * or drop the packet.
722 		 */
723 		KASSERT(vlan_input_p != NULL,("ether_input: VLAN not loaded!"));
724 		(*vlan_input_p)(ifp, m);
725 		return;
726 	}
727 
728 	/*
729 	 * Handle protocols that expect to have the Ethernet header
730 	 * (and possibly FCS) intact.
731 	 */
732 	switch (ether_type) {
733 	case ETHERTYPE_VLAN:
734 		if (ifp->if_vlantrunk != NULL) {
735 			KASSERT(vlan_input_p,("ether_input: VLAN not loaded!"));
736 			(*vlan_input_p)(ifp, m);
737 		} else {
738 			ifp->if_noproto++;
739 			m_freem(m);
740 		}
741 		return;
742 	}
743 
744 	/* Strip off Ethernet header. */
745 	m_adj(m, ETHER_HDR_LEN);
746 
747 	/* If the CRC is still on the packet, trim it off. */
748 	if (m->m_flags & M_HASFCS) {
749 		m_adj(m, -ETHER_CRC_LEN);
750 		m->m_flags &= ~M_HASFCS;
751 	}
752 
753 	/* Reset layer specific mbuf flags to avoid confusing upper layers. */
754 	m->m_flags &= ~(M_PROTOFLAGS);
755 
756 	switch (ether_type) {
757 #ifdef INET
758 	case ETHERTYPE_IP:
759 		if ((m = ip_fastforward(m)) == NULL)
760 			return;
761 		isr = NETISR_IP;
762 		break;
763 
764 	case ETHERTYPE_ARP:
765 		if (ifp->if_flags & IFF_NOARP) {
766 			/* Discard packet if ARP is disabled on interface */
767 			m_freem(m);
768 			return;
769 		}
770 		isr = NETISR_ARP;
771 		break;
772 #endif
773 #ifdef IPX
774 	case ETHERTYPE_IPX:
775 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
776 			return;
777 		isr = NETISR_IPX;
778 		break;
779 #endif
780 #ifdef INET6
781 	case ETHERTYPE_IPV6:
782 		isr = NETISR_IPV6;
783 		break;
784 #endif
785 #ifdef NETATALK
786 	case ETHERTYPE_AT:
787 		isr = NETISR_ATALK1;
788 		break;
789 	case ETHERTYPE_AARP:
790 		isr = NETISR_AARP;
791 		break;
792 #endif /* NETATALK */
793 	default:
794 #ifdef IPX
795 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
796 			return;
797 #endif /* IPX */
798 #if defined(NETATALK)
799 		if (ether_type > ETHERMTU)
800 			goto discard;
801 		l = mtod(m, struct llc *);
802 		if (l->llc_dsap == LLC_SNAP_LSAP &&
803 		    l->llc_ssap == LLC_SNAP_LSAP &&
804 		    l->llc_control == LLC_UI) {
805 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
806 			    sizeof(at_org_code)) == 0 &&
807 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
808 				m_adj(m, LLC_SNAPFRAMELEN);
809 				isr = NETISR_ATALK2;
810 				break;
811 			}
812 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
813 			    sizeof(aarp_org_code)) == 0 &&
814 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
815 				m_adj(m, LLC_SNAPFRAMELEN);
816 				isr = NETISR_AARP;
817 				break;
818 			}
819 		}
820 #endif /* NETATALK */
821 		goto discard;
822 	}
823 	netisr_dispatch(isr, m);
824 	return;
825 
826 discard:
827 	/*
828 	 * Packet is to be discarded.  If netgraph is present,
829 	 * hand the packet to it for last chance processing;
830 	 * otherwise dispose of it.
831 	 */
832 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
833 		KASSERT(ng_ether_input_orphan_p != NULL,
834 		    ("ng_ether_input_orphan_p is NULL"));
835 		/*
836 		 * Put back the ethernet header so netgraph has a
837 		 * consistent view of inbound packets.
838 		 */
839 		M_PREPEND(m, ETHER_HDR_LEN, M_DONTWAIT);
840 		(*ng_ether_input_orphan_p)(ifp, m);
841 		return;
842 	}
843 	m_freem(m);
844 }
845 
846 /*
847  * Convert Ethernet address to printable (loggable) representation.
848  * This routine is for compatibility; it's better to just use
849  *
850  *	printf("%6D", <pointer to address>, ":");
851  *
852  * since there's no static buffer involved.
853  */
854 char *
855 ether_sprintf(const u_char *ap)
856 {
857 	static char etherbuf[18];
858 	snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
859 	return (etherbuf);
860 }
861 
862 /*
863  * Perform common duties while attaching to interface list
864  */
865 void
866 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
867 {
868 	int i;
869 	struct ifaddr *ifa;
870 	struct sockaddr_dl *sdl;
871 
872 	ifp->if_addrlen = ETHER_ADDR_LEN;
873 	ifp->if_hdrlen = ETHER_HDR_LEN;
874 	if_attach(ifp);
875 	ifp->if_mtu = ETHERMTU;
876 	ifp->if_output = ether_output;
877 	ifp->if_input = ether_input;
878 	ifp->if_resolvemulti = ether_resolvemulti;
879 	if (ifp->if_baudrate == 0)
880 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
881 	ifp->if_broadcastaddr = etherbroadcastaddr;
882 
883 	ifa = ifp->if_addr;
884 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
885 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
886 	sdl->sdl_type = IFT_ETHER;
887 	sdl->sdl_alen = ifp->if_addrlen;
888 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
889 
890 	bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
891 	if (ng_ether_attach_p != NULL)
892 		(*ng_ether_attach_p)(ifp);
893 
894 	/* Announce Ethernet MAC address if non-zero. */
895 	for (i = 0; i < ifp->if_addrlen; i++)
896 		if (lla[i] != 0)
897 			break;
898 	if (i != ifp->if_addrlen)
899 		if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
900 	if (debug_mpsafenet && (ifp->if_flags & IFF_NEEDSGIANT) != 0)
901 		if_printf(ifp, "if_start running deferred for Giant\n");
902 }
903 
904 /*
905  * Perform common duties while detaching an Ethernet interface
906  */
907 void
908 ether_ifdetach(struct ifnet *ifp)
909 {
910 	if (IFP2AC(ifp)->ac_netgraph != NULL) {
911 		KASSERT(ng_ether_detach_p != NULL,
912 		    ("ng_ether_detach_p is NULL"));
913 		(*ng_ether_detach_p)(ifp);
914 	}
915 
916 	bpfdetach(ifp);
917 	if_detach(ifp);
918 }
919 
920 SYSCTL_DECL(_net_link);
921 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
922 #if defined(INET) || defined(INET6)
923 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
924 	    &ether_ipfw,0,"Pass ether pkts through firewall");
925 #endif
926 
927 #if 0
928 /*
929  * This is for reference.  We have a table-driven version
930  * of the little-endian crc32 generator, which is faster
931  * than the double-loop.
932  */
933 uint32_t
934 ether_crc32_le(const uint8_t *buf, size_t len)
935 {
936 	size_t i;
937 	uint32_t crc;
938 	int bit;
939 	uint8_t data;
940 
941 	crc = 0xffffffff;	/* initial value */
942 
943 	for (i = 0; i < len; i++) {
944 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1)
945 			carry = (crc ^ data) & 1;
946 			crc >>= 1;
947 			if (carry)
948 				crc = (crc ^ ETHER_CRC_POLY_LE);
949 	}
950 
951 	return (crc);
952 }
953 #else
954 uint32_t
955 ether_crc32_le(const uint8_t *buf, size_t len)
956 {
957 	static const uint32_t crctab[] = {
958 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
959 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
960 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
961 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
962 	};
963 	size_t i;
964 	uint32_t crc;
965 
966 	crc = 0xffffffff;	/* initial value */
967 
968 	for (i = 0; i < len; i++) {
969 		crc ^= buf[i];
970 		crc = (crc >> 4) ^ crctab[crc & 0xf];
971 		crc = (crc >> 4) ^ crctab[crc & 0xf];
972 	}
973 
974 	return (crc);
975 }
976 #endif
977 
978 uint32_t
979 ether_crc32_be(const uint8_t *buf, size_t len)
980 {
981 	size_t i;
982 	uint32_t crc, carry;
983 	int bit;
984 	uint8_t data;
985 
986 	crc = 0xffffffff;	/* initial value */
987 
988 	for (i = 0; i < len; i++) {
989 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
990 			carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
991 			crc <<= 1;
992 			if (carry)
993 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
994 		}
995 	}
996 
997 	return (crc);
998 }
999 
1000 int
1001 ether_ioctl(struct ifnet *ifp, int command, caddr_t data)
1002 {
1003 	struct ifaddr *ifa = (struct ifaddr *) data;
1004 	struct ifreq *ifr = (struct ifreq *) data;
1005 	int error = 0;
1006 
1007 	switch (command) {
1008 	case SIOCSIFADDR:
1009 		ifp->if_flags |= IFF_UP;
1010 
1011 		switch (ifa->ifa_addr->sa_family) {
1012 #ifdef INET
1013 		case AF_INET:
1014 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
1015 			arp_ifinit(ifp, ifa);
1016 			break;
1017 #endif
1018 #ifdef IPX
1019 		/*
1020 		 * XXX - This code is probably wrong
1021 		 */
1022 		case AF_IPX:
1023 			{
1024 			struct ipx_addr *ina = &(IA_SIPX(ifa)->sipx_addr);
1025 
1026 			if (ipx_nullhost(*ina))
1027 				ina->x_host =
1028 				    *(union ipx_host *)
1029 				    IF_LLADDR(ifp);
1030 			else {
1031 				bcopy((caddr_t) ina->x_host.c_host,
1032 				      (caddr_t) IF_LLADDR(ifp),
1033 				      ETHER_ADDR_LEN);
1034 			}
1035 
1036 			/*
1037 			 * Set new address
1038 			 */
1039 			ifp->if_init(ifp->if_softc);
1040 			break;
1041 			}
1042 #endif
1043 		default:
1044 			ifp->if_init(ifp->if_softc);
1045 			break;
1046 		}
1047 		break;
1048 
1049 	case SIOCGIFADDR:
1050 		{
1051 			struct sockaddr *sa;
1052 
1053 			sa = (struct sockaddr *) & ifr->ifr_data;
1054 			bcopy(IF_LLADDR(ifp),
1055 			      (caddr_t) sa->sa_data, ETHER_ADDR_LEN);
1056 		}
1057 		break;
1058 
1059 	case SIOCSIFMTU:
1060 		/*
1061 		 * Set the interface MTU.
1062 		 */
1063 		if (ifr->ifr_mtu > ETHERMTU) {
1064 			error = EINVAL;
1065 		} else {
1066 			ifp->if_mtu = ifr->ifr_mtu;
1067 		}
1068 		break;
1069 	default:
1070 		error = EINVAL;			/* XXX netbsd has ENOTTY??? */
1071 		break;
1072 	}
1073 	return (error);
1074 }
1075 
1076 static int
1077 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1078 	struct sockaddr *sa)
1079 {
1080 	struct sockaddr_dl *sdl;
1081 #ifdef INET
1082 	struct sockaddr_in *sin;
1083 #endif
1084 #ifdef INET6
1085 	struct sockaddr_in6 *sin6;
1086 #endif
1087 	u_char *e_addr;
1088 
1089 	switch(sa->sa_family) {
1090 	case AF_LINK:
1091 		/*
1092 		 * No mapping needed. Just check that it's a valid MC address.
1093 		 */
1094 		sdl = (struct sockaddr_dl *)sa;
1095 		e_addr = LLADDR(sdl);
1096 		if (!ETHER_IS_MULTICAST(e_addr))
1097 			return EADDRNOTAVAIL;
1098 		*llsa = 0;
1099 		return 0;
1100 
1101 #ifdef INET
1102 	case AF_INET:
1103 		sin = (struct sockaddr_in *)sa;
1104 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1105 			return EADDRNOTAVAIL;
1106 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
1107 		       M_NOWAIT|M_ZERO);
1108 		if (sdl == NULL)
1109 			return ENOMEM;
1110 		sdl->sdl_len = sizeof *sdl;
1111 		sdl->sdl_family = AF_LINK;
1112 		sdl->sdl_index = ifp->if_index;
1113 		sdl->sdl_type = IFT_ETHER;
1114 		sdl->sdl_alen = ETHER_ADDR_LEN;
1115 		e_addr = LLADDR(sdl);
1116 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1117 		*llsa = (struct sockaddr *)sdl;
1118 		return 0;
1119 #endif
1120 #ifdef INET6
1121 	case AF_INET6:
1122 		sin6 = (struct sockaddr_in6 *)sa;
1123 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1124 			/*
1125 			 * An IP6 address of 0 means listen to all
1126 			 * of the Ethernet multicast address used for IP6.
1127 			 * (This is used for multicast routers.)
1128 			 */
1129 			ifp->if_flags |= IFF_ALLMULTI;
1130 			*llsa = 0;
1131 			return 0;
1132 		}
1133 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1134 			return EADDRNOTAVAIL;
1135 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
1136 		       M_NOWAIT|M_ZERO);
1137 		if (sdl == NULL)
1138 			return (ENOMEM);
1139 		sdl->sdl_len = sizeof *sdl;
1140 		sdl->sdl_family = AF_LINK;
1141 		sdl->sdl_index = ifp->if_index;
1142 		sdl->sdl_type = IFT_ETHER;
1143 		sdl->sdl_alen = ETHER_ADDR_LEN;
1144 		e_addr = LLADDR(sdl);
1145 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1146 		*llsa = (struct sockaddr *)sdl;
1147 		return 0;
1148 #endif
1149 
1150 	default:
1151 		/*
1152 		 * Well, the text isn't quite right, but it's the name
1153 		 * that counts...
1154 		 */
1155 		return EAFNOSUPPORT;
1156 	}
1157 }
1158 
1159 static void*
1160 ether_alloc(u_char type, struct ifnet *ifp)
1161 {
1162 	struct arpcom	*ac;
1163 
1164 	ac = malloc(sizeof(struct arpcom), M_ARPCOM, M_WAITOK | M_ZERO);
1165 	ac->ac_ifp = ifp;
1166 
1167 	return (ac);
1168 }
1169 
1170 static void
1171 ether_free(void *com, u_char type)
1172 {
1173 
1174 	free(com, M_ARPCOM);
1175 }
1176 
1177 static int
1178 ether_modevent(module_t mod, int type, void *data)
1179 {
1180 
1181 	switch (type) {
1182 	case MOD_LOAD:
1183 		if_register_com_alloc(IFT_ETHER, ether_alloc, ether_free);
1184 		break;
1185 	case MOD_UNLOAD:
1186 		if_deregister_com_alloc(IFT_ETHER);
1187 		break;
1188 	default:
1189 		return EOPNOTSUPP;
1190 	}
1191 
1192 	return (0);
1193 }
1194 
1195 static moduledata_t ether_mod = {
1196 	"ether",
1197 	ether_modevent,
1198 	0
1199 };
1200 
1201 void
1202 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1203 {
1204 	struct ether_vlan_header vlan;
1205 	struct mbuf mv, mb;
1206 
1207 	KASSERT((m->m_flags & M_VLANTAG) != 0,
1208 	    ("%s: vlan information not present", __func__));
1209 	KASSERT(m->m_len >= sizeof(struct ether_header),
1210 	    ("%s: mbuf not large enough for header", __func__));
1211 	bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1212 	vlan.evl_proto = vlan.evl_encap_proto;
1213 	vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1214 	vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1215 	m->m_len -= sizeof(struct ether_header);
1216 	m->m_data += sizeof(struct ether_header);
1217 	/*
1218 	 * If a data link has been supplied by the caller, then we will need to
1219 	 * re-create a stack allocated mbuf chain with the following structure:
1220 	 *
1221 	 * (1) mbuf #1 will contain the supplied data link
1222 	 * (2) mbuf #2 will contain the vlan header
1223 	 * (3) mbuf #3 will contain the original mbuf's packet data
1224 	 *
1225 	 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1226 	 */
1227 	if (data != NULL) {
1228 		mv.m_next = m;
1229 		mv.m_data = (caddr_t)&vlan;
1230 		mv.m_len = sizeof(vlan);
1231 		mb.m_next = &mv;
1232 		mb.m_data = data;
1233 		mb.m_len = dlen;
1234 		bpf_mtap(bp, &mb);
1235 	} else
1236 		bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1237 	m->m_len += sizeof(struct ether_header);
1238 	m->m_data -= sizeof(struct ether_header);
1239 }
1240 
1241 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1242 MODULE_VERSION(ether, 1);
1243