xref: /netbsd-src/sys/net/if_ethersubr.c (revision 796c32c94f6e154afc9de0f63da35c91bb739b45)
1 /*	$NetBSD: if_ethersubr.c,v 1.247 2017/11/22 04:27:57 msaitoh Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (c) 1982, 1989, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)if_ethersubr.c	8.2 (Berkeley) 4/4/96
61  */
62 
63 #include <sys/cdefs.h>
64 __KERNEL_RCSID(0, "$NetBSD: if_ethersubr.c,v 1.247 2017/11/22 04:27:57 msaitoh Exp $");
65 
66 #ifdef _KERNEL_OPT
67 #include "opt_inet.h"
68 #include "opt_atalk.h"
69 #include "opt_mbuftrace.h"
70 #include "opt_mpls.h"
71 #include "opt_gateway.h"
72 #include "opt_pppoe.h"
73 #include "opt_net_mpsafe.h"
74 #endif
75 
76 #include "vlan.h"
77 #include "pppoe.h"
78 #include "bridge.h"
79 #include "arp.h"
80 #include "agr.h"
81 
82 #include <sys/sysctl.h>
83 #include <sys/mbuf.h>
84 #include <sys/mutex.h>
85 #include <sys/ioctl.h>
86 #include <sys/errno.h>
87 #include <sys/device.h>
88 #include <sys/rnd.h>
89 #include <sys/rndsource.h>
90 #include <sys/cpu.h>
91 #include <sys/kmem.h>
92 
93 #include <net/if.h>
94 #include <net/netisr.h>
95 #include <net/route.h>
96 #include <net/if_llc.h>
97 #include <net/if_dl.h>
98 #include <net/if_types.h>
99 #include <net/pktqueue.h>
100 
101 #include <net/if_media.h>
102 #include <dev/mii/mii.h>
103 #include <dev/mii/miivar.h>
104 
105 #if NARP == 0
106 /*
107  * XXX there should really be a way to issue this warning from within config(8)
108  */
109 #error You have included NETATALK or a pseudo-device in your configuration that depends on the presence of ethernet interfaces, but have no such interfaces configured. Check if you really need pseudo-device bridge, pppoe, vlan or options NETATALK.
110 #endif
111 
112 #include <net/bpf.h>
113 
114 #include <net/if_ether.h>
115 #include <net/if_vlanvar.h>
116 
117 #if NPPPOE > 0
118 #include <net/if_pppoe.h>
119 #endif
120 
121 #if NAGR > 0
122 #include <net/agr/ieee8023_slowprotocols.h>	/* XXX */
123 #include <net/agr/ieee8023ad.h>
124 #include <net/agr/if_agrvar.h>
125 #endif
126 
127 #if NBRIDGE > 0
128 #include <net/if_bridgevar.h>
129 #endif
130 
131 #include <netinet/in.h>
132 #ifdef INET
133 #include <netinet/in_var.h>
134 #endif
135 #include <netinet/if_inarp.h>
136 
137 #ifdef INET6
138 #ifndef INET
139 #include <netinet/in.h>
140 #endif
141 #include <netinet6/in6_var.h>
142 #include <netinet6/nd6.h>
143 #endif
144 
145 
146 #include "carp.h"
147 #if NCARP > 0
148 #include <netinet/ip_carp.h>
149 #endif
150 
151 #ifdef NETATALK
152 #include <netatalk/at.h>
153 #include <netatalk/at_var.h>
154 #include <netatalk/at_extern.h>
155 
156 #define llc_snap_org_code llc_un.type_snap.org_code
157 #define llc_snap_ether_type llc_un.type_snap.ether_type
158 
159 extern u_char	at_org_code[3];
160 extern u_char	aarp_org_code[3];
161 #endif /* NETATALK */
162 
163 #ifdef MPLS
164 #include <netmpls/mpls.h>
165 #include <netmpls/mpls_var.h>
166 #endif
167 
168 static struct timeval bigpktppslim_last;
169 static int bigpktppslim = 2;	/* XXX */
170 static int bigpktpps_count;
171 static kmutex_t bigpktpps_lock __cacheline_aligned;
172 
173 
174 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] =
175     { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
176 const uint8_t ethermulticastaddr_slowprotocols[ETHER_ADDR_LEN] =
177     { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x02 };
178 #define senderr(e) { error = (e); goto bad;}
179 
180 static	int ether_output(struct ifnet *, struct mbuf *,
181 	    const struct sockaddr *, const struct rtentry *);
182 
183 /*
184  * Ethernet output routine.
185  * Encapsulate a packet of type family for the local net.
186  * Assumes that ifp is actually pointer to ethercom structure.
187  */
188 static int
189 ether_output(struct ifnet * const ifp0, struct mbuf * const m0,
190 	const struct sockaddr * const dst,
191 	const struct rtentry *rt)
192 {
193 	uint16_t etype = 0;
194 	int error = 0, hdrcmplt = 0;
195  	uint8_t esrc[6], edst[6];
196 	struct mbuf *m = m0;
197 	struct mbuf *mcopy = NULL;
198 	struct ether_header *eh;
199 	struct ifnet *ifp = ifp0;
200 #ifdef INET
201 	struct arphdr *ah;
202 #endif /* INET */
203 #ifdef NETATALK
204 	struct at_ifaddr *aa;
205 #endif /* NETATALK */
206 
207 #ifdef MBUFTRACE
208 	m_claimm(m, ifp->if_mowner);
209 #endif
210 
211 #if NCARP > 0
212 	if (ifp->if_type == IFT_CARP) {
213 		struct ifaddr *ifa;
214 		int s = pserialize_read_enter();
215 
216 		/* loop back if this is going to the carp interface */
217 		if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP &&
218 		    (ifa = ifa_ifwithaddr(dst)) != NULL) {
219 			if (ifa->ifa_ifp == ifp0) {
220 				pserialize_read_exit(s);
221 				return looutput(ifp0, m, dst, rt);
222 			}
223 		}
224 		pserialize_read_exit(s);
225 
226 		ifp = ifp->if_carpdev;
227 		/* ac = (struct arpcom *)ifp; */
228 
229 		if ((ifp0->if_flags & (IFF_UP | IFF_RUNNING)) !=
230 		    (IFF_UP | IFF_RUNNING))
231 			senderr(ENETDOWN);
232 	}
233 #endif /* NCARP > 0 */
234 
235 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
236 		senderr(ENETDOWN);
237 
238 	switch (dst->sa_family) {
239 
240 #ifdef INET
241 	case AF_INET:
242 		if (m->m_flags & M_BCAST)
243 			(void)memcpy(edst, etherbroadcastaddr, sizeof(edst));
244 		else if (m->m_flags & M_MCAST)
245 			ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst);
246 		else if ((error = arpresolve(ifp, rt, m, dst, edst,
247 		    sizeof(edst))) != 0) {
248 			return error == EWOULDBLOCK ? 0 : error;
249 		}
250 		/* If broadcasting on a simplex interface, loopback a copy */
251 		if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
252 			mcopy = m_copy(m, 0, (int)M_COPYALL);
253 		etype = htons(ETHERTYPE_IP);
254 		break;
255 
256 	case AF_ARP:
257 		ah = mtod(m, struct arphdr *);
258 		if (m->m_flags & M_BCAST)
259 			(void)memcpy(edst, etherbroadcastaddr, sizeof(edst));
260 		else {
261 			void *tha = ar_tha(ah);
262 
263 			if (tha == NULL) {
264 				/* fake with ARPHDR_IEEE1394 */
265 				m_freem(m);
266 				return 0;
267 			}
268 			memcpy(edst, tha, sizeof(edst));
269 		}
270 
271 		ah->ar_hrd = htons(ARPHRD_ETHER);
272 
273 		switch (ntohs(ah->ar_op)) {
274 		case ARPOP_REVREQUEST:
275 		case ARPOP_REVREPLY:
276 			etype = htons(ETHERTYPE_REVARP);
277 			break;
278 
279 		case ARPOP_REQUEST:
280 		case ARPOP_REPLY:
281 		default:
282 			etype = htons(ETHERTYPE_ARP);
283 		}
284 
285 		break;
286 #endif
287 #ifdef INET6
288 	case AF_INET6:
289 		if (m->m_flags & M_BCAST)
290 			(void)memcpy(edst, etherbroadcastaddr, sizeof(edst));
291 		else if (m->m_flags & M_MCAST) {
292 			ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr,
293 			    edst);
294 		} else {
295 			error = nd6_resolve(ifp, rt, m, dst, edst,
296 			    sizeof(edst));
297 			if (error != 0)
298 				return error == EWOULDBLOCK ? 0 : error;
299 		}
300 		etype = htons(ETHERTYPE_IPV6);
301 		break;
302 #endif
303 #ifdef NETATALK
304     case AF_APPLETALK: {
305 		struct ifaddr *ifa;
306 		int s;
307 
308 		KERNEL_LOCK(1, NULL);
309 		if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) {
310 #ifdef NETATALKDEBUG
311 			printf("aarpresolv failed\n");
312 #endif /* NETATALKDEBUG */
313 			KERNEL_UNLOCK_ONE(NULL);
314 			return (0);
315 		}
316 		/*
317 		 * ifaddr is the first thing in at_ifaddr
318 		 */
319 		s = pserialize_read_enter();
320 		ifa = at_ifawithnet((const struct sockaddr_at *)dst, ifp);
321 		if (ifa == NULL) {
322 			pserialize_read_exit(s);
323 			KERNEL_UNLOCK_ONE(NULL);
324 			goto bad;
325 		}
326 		aa = (struct at_ifaddr *)ifa;
327 
328 		/*
329 		 * In the phase 2 case, we need to prepend an mbuf for the
330 		 * llc header.  Since we must preserve the value of m,
331 		 * which is passed to us by value, we m_copy() the first
332 		 * mbuf, and use it for our llc header.
333 		 */
334 		if (aa->aa_flags & AFA_PHASE2) {
335 			struct llc llc;
336 
337 			M_PREPEND(m, sizeof(struct llc), M_DONTWAIT);
338 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
339 			llc.llc_control = LLC_UI;
340 			memcpy(llc.llc_snap_org_code, at_org_code,
341 			    sizeof(llc.llc_snap_org_code));
342 			llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK);
343 			memcpy(mtod(m, void *), &llc, sizeof(struct llc));
344 		} else {
345 			etype = htons(ETHERTYPE_ATALK);
346 		}
347 		pserialize_read_exit(s);
348 		KERNEL_UNLOCK_ONE(NULL);
349 		break;
350 	    }
351 #endif /* NETATALK */
352 	case pseudo_AF_HDRCMPLT:
353 		hdrcmplt = 1;
354 		memcpy(esrc,
355 		    ((const struct ether_header *)dst->sa_data)->ether_shost,
356 		    sizeof(esrc));
357 		/* FALLTHROUGH */
358 
359 	case AF_UNSPEC:
360  		memcpy(edst,
361 		    ((const struct ether_header *)dst->sa_data)->ether_dhost,
362 		    sizeof(edst));
363 		/* AF_UNSPEC doesn't swap the byte order of the ether_type. */
364 		etype = ((const struct ether_header *)dst->sa_data)->ether_type;
365 		break;
366 
367 	default:
368 		printf("%s: can't handle af%d\n", ifp->if_xname,
369 			dst->sa_family);
370 		senderr(EAFNOSUPPORT);
371 	}
372 
373 #ifdef MPLS
374 	KERNEL_LOCK(1, NULL);
375 	{
376 		struct m_tag *mtag;
377 		mtag = m_tag_find(m, PACKET_TAG_MPLS, NULL);
378 		if (mtag != NULL) {
379 			/* Having the tag itself indicates it's MPLS */
380 			etype = htons(ETHERTYPE_MPLS);
381 			m_tag_delete(m, mtag);
382 		}
383 	}
384 	KERNEL_UNLOCK_ONE(NULL);
385 #endif
386 
387 	if (mcopy)
388 		(void)looutput(ifp, mcopy, dst, rt);
389 
390 	/* If no ether type is set, this must be a 802.2 formatted packet.
391 	 */
392 	if (etype == 0)
393 		etype = htons(m->m_pkthdr.len);
394 	/*
395 	 * Add local net header.  If no space in first mbuf,
396 	 * allocate another.
397 	 */
398 	M_PREPEND(m, sizeof (struct ether_header), M_DONTWAIT);
399 	if (m == 0)
400 		senderr(ENOBUFS);
401 	eh = mtod(m, struct ether_header *);
402 	/* Note: etype is already in network byte order. */
403 	(void)memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
404  	memcpy(eh->ether_dhost, edst, sizeof(edst));
405 	if (hdrcmplt)
406 		memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost));
407 	else
408 	 	memcpy(eh->ether_shost, CLLADDR(ifp->if_sadl),
409 		    sizeof(eh->ether_shost));
410 
411 #if NCARP > 0
412 	if (ifp0 != ifp && ifp0->if_type == IFT_CARP) {
413 	 	memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl),
414 		    sizeof(eh->ether_shost));
415 	}
416 #endif /* NCARP > 0 */
417 
418 	if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0)
419 		return (error);
420 	if (m == NULL)
421 		return (0);
422 
423 #if NBRIDGE > 0
424 	/*
425 	 * Bridges require special output handling.
426 	 */
427 	if (ifp->if_bridge)
428 		return (bridge_output(ifp, m, NULL, NULL));
429 #endif
430 
431 #if NCARP > 0
432 	if (ifp != ifp0)
433 		ifp0->if_obytes += m->m_pkthdr.len + ETHER_HDR_LEN;
434 #endif /* NCARP > 0 */
435 
436 #ifdef ALTQ
437 	KERNEL_LOCK(1, NULL);
438 	/*
439 	 * If ALTQ is enabled on the parent interface, do
440 	 * classification; the queueing discipline might not
441 	 * require classification, but might require the
442 	 * address family/header pointer in the pktattr.
443 	 */
444 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
445 		altq_etherclassify(&ifp->if_snd, m);
446 	KERNEL_UNLOCK_ONE(NULL);
447 #endif
448 	return ifq_enqueue(ifp, m);
449 
450 bad:
451 	if (m)
452 		m_freem(m);
453 	return (error);
454 }
455 
456 #ifdef ALTQ
457 /*
458  * This routine is a slight hack to allow a packet to be classified
459  * if the Ethernet headers are present.  It will go away when ALTQ's
460  * classification engine understands link headers.
461  */
462 void
463 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m)
464 {
465 	struct ether_header *eh;
466 	uint16_t ether_type;
467 	int hlen, af, hdrsize;
468 	void *hdr;
469 
470 	hlen = ETHER_HDR_LEN;
471 	eh = mtod(m, struct ether_header *);
472 
473 	ether_type = htons(eh->ether_type);
474 
475 	if (ether_type < ETHERMTU) {
476 		/* LLC/SNAP */
477 		struct llc *llc = (struct llc *)(eh + 1);
478 		hlen += 8;
479 
480 		if (m->m_len < hlen ||
481 		    llc->llc_dsap != LLC_SNAP_LSAP ||
482 		    llc->llc_ssap != LLC_SNAP_LSAP ||
483 		    llc->llc_control != LLC_UI) {
484 			/* Not SNAP. */
485 			goto bad;
486 		}
487 
488 		ether_type = htons(llc->llc_un.type_snap.ether_type);
489 	}
490 
491 	switch (ether_type) {
492 	case ETHERTYPE_IP:
493 		af = AF_INET;
494 		hdrsize = 20;		/* sizeof(struct ip) */
495 		break;
496 
497 	case ETHERTYPE_IPV6:
498 		af = AF_INET6;
499 		hdrsize = 40;		/* sizeof(struct ip6_hdr) */
500 		break;
501 
502 	default:
503 		af = AF_UNSPEC;
504 		hdrsize = 0;
505 		break;
506 	}
507 
508 	while (m->m_len <= hlen) {
509 		hlen -= m->m_len;
510 		m = m->m_next;
511 	}
512 	if (m->m_len < (hlen + hdrsize)) {
513 		/*
514 		 * protocol header not in a single mbuf.
515 		 * We can't cope with this situation right
516 		 * now (but it shouldn't ever happen, really, anyhow).
517 		 */
518 #ifdef DEBUG
519 		printf("altq_etherclassify: headers span multiple mbufs: "
520 		    "%d < %d\n", m->m_len, (hlen + hdrsize));
521 #endif
522 		goto bad;
523 	}
524 
525 	m->m_data += hlen;
526 	m->m_len -= hlen;
527 
528 	hdr = mtod(m, void *);
529 
530 	if (ALTQ_NEEDS_CLASSIFY(ifq))
531 		m->m_pkthdr.pattr_class =
532 		    (*ifq->altq_classify)(ifq->altq_clfier, m, af);
533 	m->m_pkthdr.pattr_af = af;
534 	m->m_pkthdr.pattr_hdr = hdr;
535 
536 	m->m_data -= hlen;
537 	m->m_len += hlen;
538 
539 	return;
540 
541  bad:
542 	m->m_pkthdr.pattr_class = NULL;
543 	m->m_pkthdr.pattr_hdr = NULL;
544 	m->m_pkthdr.pattr_af = AF_UNSPEC;
545 }
546 #endif /* ALTQ */
547 
548 /*
549  * Process a received Ethernet packet;
550  * the packet is in the mbuf chain m with
551  * the ether header.
552  */
553 void
554 ether_input(struct ifnet *ifp, struct mbuf *m)
555 {
556 	struct ethercom *ec = (struct ethercom *) ifp;
557 	pktqueue_t *pktq = NULL;
558 	struct ifqueue *inq = NULL;
559 	uint16_t etype;
560 	struct ether_header *eh;
561 	size_t ehlen;
562 	static int earlypkts;
563 	int isr = 0;
564 #if defined (LLC) || defined(NETATALK)
565 	struct llc *l;
566 #endif
567 
568 	KASSERT(!cpu_intr_p());
569 
570 	if ((ifp->if_flags & IFF_UP) == 0) {
571 		m_freem(m);
572 		return;
573 	}
574 
575 #ifdef MBUFTRACE
576 	m_claimm(m, &ec->ec_rx_mowner);
577 #endif
578 	eh = mtod(m, struct ether_header *);
579 	etype = ntohs(eh->ether_type);
580 	ehlen = sizeof(*eh);
581 
582 	if(__predict_false(earlypkts < 100 || !rnd_initial_entropy)) {
583 		rnd_add_data(NULL, eh, ehlen, 0);
584 		earlypkts++;
585 	}
586 
587 	/*
588 	 * Determine if the packet is within its size limits.
589 	 */
590 	if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len >
591 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
592 		mutex_enter(&bigpktpps_lock);
593 		if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count,
594 			    bigpktppslim)) {
595 			printf("%s: discarding oversize frame (len=%d)\n",
596 			    ifp->if_xname, m->m_pkthdr.len);
597 		}
598 		mutex_exit(&bigpktpps_lock);
599 		m_freem(m);
600 		return;
601 	}
602 
603 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
604 		/*
605 		 * If this is not a simplex interface, drop the packet
606 		 * if it came from us.
607 		 */
608 		if ((ifp->if_flags & IFF_SIMPLEX) == 0 &&
609 		    memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost,
610 		    ETHER_ADDR_LEN) == 0) {
611 			m_freem(m);
612 			return;
613 		}
614 
615 		if (memcmp(etherbroadcastaddr,
616 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
617 			m->m_flags |= M_BCAST;
618 		else
619 			m->m_flags |= M_MCAST;
620 		ifp->if_imcasts++;
621 	}
622 
623 	/* If the CRC is still on the packet, trim it off. */
624 	if (m->m_flags & M_HASFCS) {
625 		m_adj(m, -ETHER_CRC_LEN);
626 		m->m_flags &= ~M_HASFCS;
627 	}
628 
629 	ifp->if_ibytes += m->m_pkthdr.len;
630 
631 #if NCARP > 0
632 	if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) {
633 		/*
634 		 * clear M_PROMISC, in case the packets comes from a
635 		 * vlan
636 		 */
637 		m->m_flags &= ~M_PROMISC;
638 		if (carp_input(m, (uint8_t *)&eh->ether_shost,
639 		    (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0)
640 			return;
641 	}
642 #endif /* NCARP > 0 */
643 	if ((m->m_flags & (M_BCAST | M_MCAST | M_PROMISC)) == 0 &&
644 	    (ifp->if_flags & IFF_PROMISC) != 0 &&
645 	    memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost,
646 		   ETHER_ADDR_LEN) != 0) {
647 		m->m_flags |= M_PROMISC;
648 	}
649 
650 	if ((m->m_flags & M_PROMISC) == 0) {
651 		if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
652 			return;
653 		if (m == NULL)
654 			return;
655 
656 		eh = mtod(m, struct ether_header *);
657 		etype = ntohs(eh->ether_type);
658 		ehlen = sizeof(*eh);
659 	}
660 
661 #if NAGR > 0
662 	if (ifp->if_agrprivate &&
663 	    __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) {
664 		m->m_flags &= ~M_PROMISC;
665 		agr_input(ifp, m);
666 		return;
667 	}
668 #endif /* NAGR > 0 */
669 
670 	/*
671 	 * If VLANs are configured on the interface, check to
672 	 * see if the device performed the decapsulation and
673 	 * provided us with the tag.
674 	 */
675 	if (ec->ec_nvlans && vlan_has_tag(m)) {
676 #if NVLAN > 0
677 		/*
678 		 * vlan_input() will either recursively call ether_input()
679 		 * or drop the packet.
680 		 */
681 		vlan_input(ifp, m);
682 #else
683 		m_freem(m);
684 #endif
685 		return;
686 	}
687 
688 	/*
689 	 * Handle protocols that expect to have the Ethernet header
690 	 * (and possibly FCS) intact.
691 	 */
692 	switch (etype) {
693 	case ETHERTYPE_VLAN: {
694 		struct ether_vlan_header *evl = (void *)eh;
695 		/*
696 		 * If there is a tag of 0, then the VLAN header was probably
697 		 * just being used to store the priority.  Extract the ether
698 		 * type, and if IP or IPV6, let them deal with it.
699 		 */
700 		if (m->m_len <= sizeof(*evl)
701 		    && EVL_VLANOFTAG(evl->evl_tag) == 0) {
702 			etype = ntohs(evl->evl_proto);
703 			ehlen = sizeof(*evl);
704 			if ((m->m_flags & M_PROMISC) == 0
705 			    && (etype == ETHERTYPE_IP
706 				|| etype == ETHERTYPE_IPV6))
707 				break;
708 		}
709 #if NVLAN > 0
710 		/*
711 		 * vlan_input() will either recursively call ether_input()
712 		 * or drop the packet.
713 		 */
714 		if (((struct ethercom *)ifp)->ec_nvlans != 0)
715 			vlan_input(ifp, m);
716 		else
717 #endif /* NVLAN > 0 */
718 			m_freem(m);
719 		return;
720 	}
721 #if NPPPOE > 0
722 	case ETHERTYPE_PPPOEDISC:
723 		pppoedisc_input(ifp, m);
724 		return;
725 	case ETHERTYPE_PPPOE:
726 		pppoe_input(ifp, m);
727 		return;
728 #endif /* NPPPOE > 0 */
729 	case ETHERTYPE_SLOWPROTOCOLS: {
730 		uint8_t subtype;
731 
732 #if defined(DIAGNOSTIC)
733 		if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype)) {
734 			panic("ether_input: too short slow protocol packet");
735 		}
736 #endif
737 		m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype);
738 		switch (subtype) {
739 #if NAGR > 0
740 		case SLOWPROTOCOLS_SUBTYPE_LACP:
741 			if (ifp->if_agrprivate) {
742 				ieee8023ad_lacp_input(ifp, m);
743 				return;
744 			}
745 			break;
746 
747 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
748 			if (ifp->if_agrprivate) {
749 				ieee8023ad_marker_input(ifp, m);
750 				return;
751 			}
752 			break;
753 #endif /* NAGR > 0 */
754 		default:
755 			if (subtype == 0 || subtype > 10) {
756 				/* illegal value */
757 				m_freem(m);
758 				return;
759 			}
760 			/* unknown subtype */
761 			break;
762 		}
763 		/* FALLTHROUGH */
764 	}
765 	default:
766 		if (m->m_flags & M_PROMISC) {
767 			m_freem(m);
768 			return;
769 		}
770 	}
771 
772 	/* If the CRC is still on the packet, trim it off. */
773 	if (m->m_flags & M_HASFCS) {
774 		m_adj(m, -ETHER_CRC_LEN);
775 		m->m_flags &= ~M_HASFCS;
776 	}
777 
778 	if (etype > ETHERMTU + sizeof (struct ether_header)) {
779 		/* Strip off the Ethernet header. */
780 		m_adj(m, ehlen);
781 
782 		switch (etype) {
783 #ifdef INET
784 		case ETHERTYPE_IP:
785 #ifdef GATEWAY
786 			if (ipflow_fastforward(m))
787 				return;
788 #endif
789 			pktq = ip_pktq;
790 			break;
791 
792 		case ETHERTYPE_ARP:
793 			isr = NETISR_ARP;
794 			inq = &arpintrq;
795 			break;
796 
797 		case ETHERTYPE_REVARP:
798 			revarpinput(m);	/* XXX queue? */
799 			return;
800 #endif
801 #ifdef INET6
802 		case ETHERTYPE_IPV6:
803 			if (__predict_false(!in6_present)) {
804 				m_freem(m);
805 				return;
806 			}
807 #ifdef GATEWAY
808 			if (ip6flow_fastforward(&m))
809 				return;
810 #endif
811 			pktq = ip6_pktq;
812 			break;
813 #endif
814 #ifdef NETATALK
815 		case ETHERTYPE_ATALK:
816 			isr = NETISR_ATALK;
817 			inq = &atintrq1;
818 			break;
819 		case ETHERTYPE_AARP:
820 			/* probably this should be done with a NETISR as well */
821 			aarpinput(ifp, m); /* XXX */
822 			return;
823 #endif /* NETATALK */
824 #ifdef MPLS
825 		case ETHERTYPE_MPLS:
826 			isr = NETISR_MPLS;
827 			inq = &mplsintrq;
828 			break;
829 #endif
830 		default:
831 			m_freem(m);
832 			return;
833 		}
834 	} else {
835 #if defined (LLC) || defined (NETATALK)
836 		l = (struct llc *)(eh+1);
837 		switch (l->llc_dsap) {
838 #ifdef NETATALK
839 		case LLC_SNAP_LSAP:
840 			switch (l->llc_control) {
841 			case LLC_UI:
842 				if (l->llc_ssap != LLC_SNAP_LSAP) {
843 					goto dropanyway;
844 				}
845 
846 				if (memcmp(&(l->llc_snap_org_code)[0],
847 				    at_org_code, sizeof(at_org_code)) == 0 &&
848 				    ntohs(l->llc_snap_ether_type) ==
849 				    ETHERTYPE_ATALK) {
850 					inq = &atintrq2;
851 					m_adj(m, sizeof(struct ether_header)
852 					    + sizeof(struct llc));
853 					isr = NETISR_ATALK;
854 					break;
855 				}
856 
857 				if (memcmp(&(l->llc_snap_org_code)[0],
858 				    aarp_org_code,
859 				    sizeof(aarp_org_code)) == 0 &&
860 				    ntohs(l->llc_snap_ether_type) ==
861 				    ETHERTYPE_AARP) {
862 					m_adj( m, sizeof(struct ether_header)
863 					    + sizeof(struct llc));
864 					aarpinput(ifp, m); /* XXX */
865 				    return;
866 				}
867 
868 			default:
869 				goto dropanyway;
870 			}
871 			break;
872 		dropanyway:
873 #endif
874 		default:
875 			m_freem(m);
876 			return;
877 		}
878 #else /* ISO || LLC || NETATALK*/
879 		m_freem(m);
880 		return;
881 #endif /* ISO || LLC || NETATALK*/
882 	}
883 
884 	if (__predict_true(pktq)) {
885 #ifdef NET_MPSAFE
886 		const u_int h = curcpu()->ci_index;
887 #else
888 		const uint32_t h = pktq_rps_hash(m);
889 #endif
890 		if (__predict_false(!pktq_enqueue(pktq, m, h))) {
891 			m_freem(m);
892 		}
893 		return;
894 	}
895 
896 	if (__predict_false(!inq)) {
897 		/* Should not happen. */
898 		m_freem(m);
899 		return;
900 	}
901 
902 	IFQ_LOCK(inq);
903 	if (IF_QFULL(inq)) {
904 		IF_DROP(inq);
905 		IFQ_UNLOCK(inq);
906 		m_freem(m);
907 	} else {
908 		IF_ENQUEUE(inq, m);
909 		IFQ_UNLOCK(inq);
910 		schednetisr(isr);
911 	}
912 }
913 
914 /*
915  * Convert Ethernet address to printable (loggable) representation.
916  */
917 char *
918 ether_sprintf(const u_char *ap)
919 {
920 	static char etherbuf[3 * ETHER_ADDR_LEN];
921 	return ether_snprintf(etherbuf, sizeof(etherbuf), ap);
922 }
923 
924 char *
925 ether_snprintf(char *buf, size_t len, const u_char *ap)
926 {
927 	char *cp = buf;
928 	size_t i;
929 
930 	for (i = 0; i < len / 3; i++) {
931 		*cp++ = hexdigits[*ap >> 4];
932 		*cp++ = hexdigits[*ap++ & 0xf];
933 		*cp++ = ':';
934 	}
935 	*--cp = '\0';
936 	return buf;
937 }
938 
939 /*
940  * Perform common duties while attaching to interface list
941  */
942 void
943 ether_ifattach(struct ifnet *ifp, const uint8_t *lla)
944 {
945 	struct ethercom *ec = (struct ethercom *)ifp;
946 
947 	ifp->if_type = IFT_ETHER;
948 	ifp->if_hdrlen = ETHER_HDR_LEN;
949 	ifp->if_dlt = DLT_EN10MB;
950 	ifp->if_mtu = ETHERMTU;
951 	ifp->if_output = ether_output;
952 	ifp->_if_input = ether_input;
953 	if (ifp->if_baudrate == 0)
954 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
955 
956 	if (lla != NULL)
957 		if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla));
958 
959 	LIST_INIT(&ec->ec_multiaddrs);
960 	ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
961 	ifp->if_broadcastaddr = etherbroadcastaddr;
962 	bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header));
963 #ifdef MBUFTRACE
964 	strlcpy(ec->ec_tx_mowner.mo_name, ifp->if_xname,
965 	    sizeof(ec->ec_tx_mowner.mo_name));
966 	strlcpy(ec->ec_tx_mowner.mo_descr, "tx",
967 	    sizeof(ec->ec_tx_mowner.mo_descr));
968 	strlcpy(ec->ec_rx_mowner.mo_name, ifp->if_xname,
969 	    sizeof(ec->ec_rx_mowner.mo_name));
970 	strlcpy(ec->ec_rx_mowner.mo_descr, "rx",
971 	    sizeof(ec->ec_rx_mowner.mo_descr));
972 	MOWNER_ATTACH(&ec->ec_tx_mowner);
973 	MOWNER_ATTACH(&ec->ec_rx_mowner);
974 	ifp->if_mowner = &ec->ec_tx_mowner;
975 #endif
976 }
977 
978 void
979 ether_ifdetach(struct ifnet *ifp)
980 {
981 	struct ethercom *ec = (void *) ifp;
982 	struct ether_multi *enm;
983 
984 	/*
985 	 * Prevent further calls to ioctl (for example turning off
986 	 * promiscuous mode from the bridge code), which eventually can
987 	 * call if_init() which can cause panics because the interface
988 	 * is in the process of being detached. Return device not configured
989 	 * instead.
990 	 */
991 	ifp->if_ioctl = (int (*)(struct ifnet *, u_long, void *))enxio;
992 
993 #if NBRIDGE > 0
994 	if (ifp->if_bridge)
995 		bridge_ifdetach(ifp);
996 #endif
997 
998 	bpf_detach(ifp);
999 
1000 #if NVLAN > 0
1001 	if (ec->ec_nvlans)
1002 		vlan_ifdetach(ifp);
1003 #endif
1004 
1005 	ETHER_LOCK(ec);
1006 	while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) {
1007 		LIST_REMOVE(enm, enm_list);
1008 		kmem_intr_free(enm, sizeof(*enm));
1009 		ec->ec_multicnt--;
1010 	}
1011 	ETHER_UNLOCK(ec);
1012 
1013 	mutex_destroy(ec->ec_lock);
1014 
1015 	ifp->if_mowner = NULL;
1016 	MOWNER_DETACH(&ec->ec_rx_mowner);
1017 	MOWNER_DETACH(&ec->ec_tx_mowner);
1018 }
1019 
1020 #if 0
1021 /*
1022  * This is for reference.  We have a table-driven version
1023  * of the little-endian crc32 generator, which is faster
1024  * than the double-loop.
1025  */
1026 uint32_t
1027 ether_crc32_le(const uint8_t *buf, size_t len)
1028 {
1029 	uint32_t c, crc, carry;
1030 	size_t i, j;
1031 
1032 	crc = 0xffffffffU;	/* initial value */
1033 
1034 	for (i = 0; i < len; i++) {
1035 		c = buf[i];
1036 		for (j = 0; j < 8; j++) {
1037 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1038 			crc >>= 1;
1039 			c >>= 1;
1040 			if (carry)
1041 				crc = (crc ^ ETHER_CRC_POLY_LE);
1042 		}
1043 	}
1044 
1045 	return (crc);
1046 }
1047 #else
1048 uint32_t
1049 ether_crc32_le(const uint8_t *buf, size_t len)
1050 {
1051 	static const uint32_t crctab[] = {
1052 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1053 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1054 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1055 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1056 	};
1057 	uint32_t crc;
1058 	size_t i;
1059 
1060 	crc = 0xffffffffU;	/* initial value */
1061 
1062 	for (i = 0; i < len; i++) {
1063 		crc ^= buf[i];
1064 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1065 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1066 	}
1067 
1068 	return (crc);
1069 }
1070 #endif
1071 
1072 uint32_t
1073 ether_crc32_be(const uint8_t *buf, size_t len)
1074 {
1075 	uint32_t c, crc, carry;
1076 	size_t i, j;
1077 
1078 	crc = 0xffffffffU;	/* initial value */
1079 
1080 	for (i = 0; i < len; i++) {
1081 		c = buf[i];
1082 		for (j = 0; j < 8; j++) {
1083 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1084 			crc <<= 1;
1085 			c >>= 1;
1086 			if (carry)
1087 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1088 		}
1089 	}
1090 
1091 	return (crc);
1092 }
1093 
1094 #ifdef INET
1095 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] =
1096     { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 };
1097 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] =
1098     { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff };
1099 #endif
1100 #ifdef INET6
1101 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] =
1102     { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 };
1103 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] =
1104     { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff };
1105 #endif
1106 
1107 /*
1108  * ether_aton implementation, not using a static buffer.
1109  */
1110 int
1111 ether_aton_r(u_char *dest, size_t len, const char *str)
1112 {
1113         const u_char *cp = (const void *)str;
1114 	u_char *ep;
1115 
1116 #define atox(c)	(((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10))
1117 
1118 	if (len < ETHER_ADDR_LEN)
1119 		return ENOSPC;
1120 
1121 	ep = dest + ETHER_ADDR_LEN;
1122 
1123 	while (*cp) {
1124                 if (!isxdigit(*cp))
1125                         return EINVAL;
1126 		*dest = atox(*cp);
1127 		cp++;
1128                 if (isxdigit(*cp)) {
1129                         *dest = (*dest << 4) | atox(*cp);
1130 			dest++;
1131 			cp++;
1132                 } else
1133 			dest++;
1134 		if (dest == ep)
1135 			return *cp == '\0' ? 0 : ENAMETOOLONG;
1136 		switch (*cp) {
1137 		case ':':
1138 		case '-':
1139 		case '.':
1140 			cp++;
1141 			break;
1142 		}
1143         }
1144 	return ENOBUFS;
1145 }
1146 
1147 /*
1148  * Convert a sockaddr into an Ethernet address or range of Ethernet
1149  * addresses.
1150  */
1151 int
1152 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN],
1153     uint8_t addrhi[ETHER_ADDR_LEN])
1154 {
1155 #ifdef INET
1156 	const struct sockaddr_in *sin;
1157 #endif /* INET */
1158 #ifdef INET6
1159 	const struct sockaddr_in6 *sin6;
1160 #endif /* INET6 */
1161 
1162 	switch (sa->sa_family) {
1163 
1164 	case AF_UNSPEC:
1165 		memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN);
1166 		memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1167 		break;
1168 
1169 #ifdef INET
1170 	case AF_INET:
1171 		sin = satocsin(sa);
1172 		if (sin->sin_addr.s_addr == INADDR_ANY) {
1173 			/*
1174 			 * An IP address of INADDR_ANY means listen to
1175 			 * or stop listening to all of the Ethernet
1176 			 * multicast addresses used for IP.
1177 			 * (This is for the sake of IP multicast routers.)
1178 			 */
1179 			memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN);
1180 			memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN);
1181 		}
1182 		else {
1183 			ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo);
1184 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1185 		}
1186 		break;
1187 #endif
1188 #ifdef INET6
1189 	case AF_INET6:
1190 		sin6 = satocsin6(sa);
1191 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1192 			/*
1193 			 * An IP6 address of 0 means listen to or stop
1194 			 * listening to all of the Ethernet multicast
1195 			 * address used for IP6.
1196 			 * (This is used for multicast routers.)
1197 			 */
1198 			memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN);
1199 			memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN);
1200 		} else {
1201 			ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo);
1202 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1203 		}
1204 		break;
1205 #endif
1206 
1207 	default:
1208 		return EAFNOSUPPORT;
1209 	}
1210 	return 0;
1211 }
1212 
1213 /*
1214  * Add an Ethernet multicast address or range of addresses to the list for a
1215  * given interface.
1216  */
1217 int
1218 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec)
1219 {
1220 	struct ether_multi *enm, *_enm;
1221 	u_char addrlo[ETHER_ADDR_LEN];
1222 	u_char addrhi[ETHER_ADDR_LEN];
1223 	int error = 0;
1224 
1225 	/* Allocate out of lock */
1226 	/* XXX still can be called in softint */
1227 	enm = kmem_intr_alloc(sizeof(*enm), KM_SLEEP);
1228 	if (enm == NULL)
1229 		return ENOBUFS;
1230 
1231 	ETHER_LOCK(ec);
1232 	error = ether_multiaddr(sa, addrlo, addrhi);
1233 	if (error != 0)
1234 		goto out;
1235 
1236 	/*
1237 	 * Verify that we have valid Ethernet multicast addresses.
1238 	 */
1239 	if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) {
1240 		error = EINVAL;
1241 		goto out;
1242 	}
1243 	/*
1244 	 * See if the address range is already in the list.
1245 	 */
1246 	ETHER_LOOKUP_MULTI(addrlo, addrhi, ec, _enm);
1247 	if (_enm != NULL) {
1248 		/*
1249 		 * Found it; just increment the reference count.
1250 		 */
1251 		++_enm->enm_refcount;
1252 		error = 0;
1253 		goto out;
1254 	}
1255 	/*
1256 	 * Link a new multicast record into the interface's multicast list.
1257 	 */
1258 	memcpy(enm->enm_addrlo, addrlo, 6);
1259 	memcpy(enm->enm_addrhi, addrhi, 6);
1260 	enm->enm_refcount = 1;
1261 	LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list);
1262 	ec->ec_multicnt++;
1263 	/*
1264 	 * Return ENETRESET to inform the driver that the list has changed
1265 	 * and its reception filter should be adjusted accordingly.
1266 	 */
1267 	error = ENETRESET;
1268 	enm = NULL;
1269 out:
1270 	ETHER_UNLOCK(ec);
1271 	if (enm != NULL)
1272 		kmem_intr_free(enm, sizeof(*enm));
1273 	return error;
1274 }
1275 
1276 /*
1277  * Delete a multicast address record.
1278  */
1279 int
1280 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec)
1281 {
1282 	struct ether_multi *enm;
1283 	u_char addrlo[ETHER_ADDR_LEN];
1284 	u_char addrhi[ETHER_ADDR_LEN];
1285 	int error;
1286 
1287 	ETHER_LOCK(ec);
1288 	error = ether_multiaddr(sa, addrlo, addrhi);
1289 	if (error != 0)
1290 		goto error;
1291 
1292 	/*
1293 	 * Look ur the address in our list.
1294 	 */
1295 	ETHER_LOOKUP_MULTI(addrlo, addrhi, ec, enm);
1296 	if (enm == NULL) {
1297 		error = ENXIO;
1298 		goto error;
1299 	}
1300 	if (--enm->enm_refcount != 0) {
1301 		/*
1302 		 * Still some claims to this record.
1303 		 */
1304 		error = 0;
1305 		goto error;
1306 	}
1307 	/*
1308 	 * No remaining claims to this record; unlink and free it.
1309 	 */
1310 	LIST_REMOVE(enm, enm_list);
1311 	ec->ec_multicnt--;
1312 	ETHER_UNLOCK(ec);
1313 
1314 	kmem_intr_free(enm, sizeof(*enm));
1315 	/*
1316 	 * Return ENETRESET to inform the driver that the list has changed
1317 	 * and its reception filter should be adjusted accordingly.
1318 	 */
1319 	return ENETRESET;
1320 error:
1321 	ETHER_UNLOCK(ec);
1322 	return error;
1323 }
1324 
1325 void
1326 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb)
1327 {
1328 	ec->ec_ifflags_cb = cb;
1329 }
1330 
1331 /*
1332  * Common ioctls for Ethernet interfaces.  Note, we must be
1333  * called at splnet().
1334  */
1335 int
1336 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data)
1337 {
1338 	struct ethercom *ec = (void *) ifp;
1339 	struct eccapreq *eccr;
1340 	struct ifreq *ifr = (struct ifreq *)data;
1341 	struct if_laddrreq *iflr = data;
1342 	const struct sockaddr_dl *sdl;
1343 	static const uint8_t zero[ETHER_ADDR_LEN];
1344 	int error;
1345 
1346 	switch (cmd) {
1347 	case SIOCINITIFADDR:
1348 	    {
1349 		struct ifaddr *ifa = (struct ifaddr *)data;
1350 		if (ifa->ifa_addr->sa_family != AF_LINK
1351 		    && (ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
1352 		       (IFF_UP | IFF_RUNNING)) {
1353 			ifp->if_flags |= IFF_UP;
1354 			if ((error = (*ifp->if_init)(ifp)) != 0)
1355 				return error;
1356 		}
1357 #ifdef INET
1358 		if (ifa->ifa_addr->sa_family == AF_INET)
1359 			arp_ifinit(ifp, ifa);
1360 #endif /* INET */
1361 		return 0;
1362 	    }
1363 
1364 	case SIOCSIFMTU:
1365 	    {
1366 		int maxmtu;
1367 
1368 		if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU)
1369 			maxmtu = ETHERMTU_JUMBO;
1370 		else
1371 			maxmtu = ETHERMTU;
1372 
1373 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu)
1374 			return EINVAL;
1375 		else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET)
1376 			return error;
1377 		else if (ifp->if_flags & IFF_UP) {
1378 			/* Make sure the device notices the MTU change. */
1379 			return (*ifp->if_init)(ifp);
1380 		} else
1381 			return 0;
1382 	    }
1383 
1384 	case SIOCSIFFLAGS:
1385 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
1386 			return error;
1387 		switch (ifp->if_flags & (IFF_UP | IFF_RUNNING)) {
1388 		case IFF_RUNNING:
1389 			/*
1390 			 * If interface is marked down and it is running,
1391 			 * then stop and disable it.
1392 			 */
1393 			(*ifp->if_stop)(ifp, 1);
1394 			break;
1395 		case IFF_UP:
1396 			/*
1397 			 * If interface is marked up and it is stopped, then
1398 			 * start it.
1399 			 */
1400 			return (*ifp->if_init)(ifp);
1401 		case IFF_UP | IFF_RUNNING:
1402 			error = 0;
1403 			if (ec->ec_ifflags_cb != NULL) {
1404 				error = (*ec->ec_ifflags_cb)(ec);
1405 				if (error == ENETRESET) {
1406 					/*
1407 					 * Reset the interface to pick up
1408 					 * changes in any other flags that
1409 					 * affect the hardware state.
1410 					 */
1411 					return (*ifp->if_init)(ifp);
1412 				}
1413 			} else
1414 				error = (*ifp->if_init)(ifp);
1415 			return error;
1416 		case 0:
1417 			break;
1418 		}
1419 		return 0;
1420 	case SIOCGETHERCAP:
1421 		eccr = (struct eccapreq *)data;
1422 		eccr->eccr_capabilities = ec->ec_capabilities;
1423 		eccr->eccr_capenable = ec->ec_capenable;
1424 		return 0;
1425 	case SIOCADDMULTI:
1426 		return ether_addmulti(ifreq_getaddr(cmd, ifr), ec);
1427 	case SIOCDELMULTI:
1428 		return ether_delmulti(ifreq_getaddr(cmd, ifr), ec);
1429 	case SIOCSIFMEDIA:
1430 	case SIOCGIFMEDIA:
1431 		if (ec->ec_mii == NULL)
1432 			return ENOTTY;
1433 		return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media, cmd);
1434 	case SIOCALIFADDR:
1435 		sdl = satocsdl(sstocsa(&iflr->addr));
1436 		if (sdl->sdl_family != AF_LINK)
1437 			;
1438 		else if (ETHER_IS_MULTICAST(CLLADDR(sdl)))
1439 			return EINVAL;
1440 		else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0)
1441 			return EINVAL;
1442 		/*FALLTHROUGH*/
1443 	default:
1444 		return ifioctl_common(ifp, cmd, data);
1445 	}
1446 	return 0;
1447 }
1448 
1449 /*
1450  * Enable/disable passing VLAN packets if the parent interface supports it.
1451  * Return:
1452  * 	 0: Ok
1453  *	-1: Parent interface does not support vlans
1454  *	>0: Error
1455  */
1456 int
1457 ether_enable_vlan_mtu(struct ifnet *ifp)
1458 {
1459 	int error;
1460 	struct ethercom *ec = (void *)ifp;
1461 
1462 	/* Parent does not support VLAN's */
1463 	if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0)
1464 		return -1;
1465 
1466 	/*
1467 	 * Parent supports the VLAN_MTU capability,
1468 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames;
1469 	 * enable it.
1470 	 */
1471 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1472 
1473 	/* Interface is down, defer for later */
1474 	if ((ifp->if_flags & IFF_UP) == 0)
1475 		return 0;
1476 
1477 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1478 		return 0;
1479 
1480 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1481 	return error;
1482 }
1483 
1484 int
1485 ether_disable_vlan_mtu(struct ifnet *ifp)
1486 {
1487 	int error;
1488 	struct ethercom *ec = (void *)ifp;
1489 
1490 	/* We still have VLAN's, defer for later */
1491 	if (ec->ec_nvlans != 0)
1492 		return 0;
1493 
1494 	/* Parent does not support VLAB's, nothing to do. */
1495 	if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0)
1496 		return -1;
1497 
1498 	/*
1499 	 * Disable Tx/Rx of VLAN-sized frames.
1500 	 */
1501 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1502 
1503 	/* Interface is down, defer for later */
1504 	if ((ifp->if_flags & IFF_UP) == 0)
1505 		return 0;
1506 
1507 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1508 		return 0;
1509 
1510 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1511 	return error;
1512 }
1513 
1514 static int
1515 ether_multicast_sysctl(SYSCTLFN_ARGS)
1516 {
1517 	struct ether_multi *enm;
1518 	struct ifnet *ifp;
1519 	struct ethercom *ec;
1520 	int error = 0;
1521 	size_t written;
1522 	struct psref psref;
1523 	int bound;
1524 	unsigned int multicnt;
1525 	struct ether_multi_sysctl *addrs;
1526 	int i;
1527 
1528 	if (namelen != 1)
1529 		return EINVAL;
1530 
1531 	bound = curlwp_bind();
1532 	ifp = if_get_byindex(name[0], &psref);
1533 	if (ifp == NULL) {
1534 		error = ENODEV;
1535 		goto out;
1536 	}
1537 	if (ifp->if_type != IFT_ETHER) {
1538 		if_put(ifp, &psref);
1539 		*oldlenp = 0;
1540 		goto out;
1541 	}
1542 	ec = (struct ethercom *)ifp;
1543 
1544 	if (oldp == NULL) {
1545 		if_put(ifp, &psref);
1546 		*oldlenp = ec->ec_multicnt * sizeof(*addrs);
1547 		goto out;
1548 	}
1549 
1550 	/*
1551 	 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which
1552 	 * is sleepable, with holding it. Copy data to a local buffer first
1553 	 * with holding it and then call sysctl_copyout without holding it.
1554 	 */
1555 retry:
1556 	multicnt = ec->ec_multicnt;
1557 	addrs = kmem_alloc(sizeof(*addrs) * multicnt, KM_SLEEP);
1558 
1559 	ETHER_LOCK(ec);
1560 	if (multicnt < ec->ec_multicnt) {
1561 		/* The number of multicast addresses have increased */
1562 		ETHER_UNLOCK(ec);
1563 		kmem_free(addrs, sizeof(*addrs) * multicnt);
1564 		goto retry;
1565 	}
1566 
1567 	i = 0;
1568 	LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) {
1569 		struct ether_multi_sysctl *addr = &addrs[i];
1570 		addr->enm_refcount = enm->enm_refcount;
1571 		memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN);
1572 		memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN);
1573 		i++;
1574 	}
1575 	ETHER_UNLOCK(ec);
1576 
1577 	error = 0;
1578 	written = 0;
1579 	for (i = 0; i < multicnt; i++) {
1580 		struct ether_multi_sysctl *addr = &addrs[i];
1581 
1582 		if (written + sizeof(*addr) > *oldlenp)
1583 			break;
1584 		error = sysctl_copyout(l, addr, oldp, sizeof(*addr));
1585 		if (error)
1586 			break;
1587 		written += sizeof(*addr);
1588 		oldp = (char *)oldp + sizeof(*addr);
1589 	}
1590 	kmem_free(addrs, sizeof(*addrs) * multicnt);
1591 
1592 	if_put(ifp, &psref);
1593 
1594 	*oldlenp = written;
1595 out:
1596 	curlwp_bindx(bound);
1597 	return error;
1598 }
1599 
1600 static void
1601 ether_sysctl_setup(struct sysctllog **clog)
1602 {
1603 	const struct sysctlnode *rnode = NULL;
1604 
1605 	sysctl_createv(clog, 0, NULL, &rnode,
1606 		       CTLFLAG_PERMANENT,
1607 		       CTLTYPE_NODE, "ether",
1608 		       SYSCTL_DESCR("Ethernet-specific information"),
1609 		       NULL, 0, NULL, 0,
1610 		       CTL_NET, CTL_CREATE, CTL_EOL);
1611 
1612 	sysctl_createv(clog, 0, &rnode, NULL,
1613 		       CTLFLAG_PERMANENT,
1614 		       CTLTYPE_NODE, "multicast",
1615 		       SYSCTL_DESCR("multicast addresses"),
1616 		       ether_multicast_sysctl, 0, NULL, 0,
1617 		       CTL_CREATE, CTL_EOL);
1618 }
1619 
1620 void
1621 etherinit(void)
1622 {
1623 
1624 	mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET);
1625 	ether_sysctl_setup(NULL);
1626 }
1627