xref: /netbsd-src/sys/net/if_ethersubr.c (revision c9496f6b604074a9451a67df576a5b423068e71e)
1 /*	$NetBSD: if_ethersubr.c,v 1.257 2018/01/19 12:31:27 nakayama 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 1.257 2018/01/19 12:31:27 nakayama 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, const struct rtentry *rt)
191 {
192 	uint8_t esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
193 	uint16_t etype = 0;
194 	int error = 0, hdrcmplt = 0;
195 	struct mbuf *m = m0;
196 	struct mbuf *mcopy = NULL;
197 	struct ether_header *eh;
198 	struct ifnet *ifp = ifp0;
199 #ifdef INET
200 	struct arphdr *ah;
201 #endif /* INET */
202 #ifdef NETATALK
203 	struct at_ifaddr *aa;
204 #endif /* NETATALK */
205 
206 #ifdef MBUFTRACE
207 	m_claimm(m, ifp->if_mowner);
208 #endif
209 
210 #if NCARP > 0
211 	if (ifp->if_type == IFT_CARP) {
212 		struct ifaddr *ifa;
213 		int s = pserialize_read_enter();
214 
215 		/* loop back if this is going to the carp interface */
216 		if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP &&
217 		    (ifa = ifa_ifwithaddr(dst)) != NULL) {
218 			if (ifa->ifa_ifp == ifp0) {
219 				pserialize_read_exit(s);
220 				return looutput(ifp0, m, dst, rt);
221 			}
222 		}
223 		pserialize_read_exit(s);
224 
225 		ifp = ifp->if_carpdev;
226 		/* ac = (struct arpcom *)ifp; */
227 
228 		if ((ifp0->if_flags & (IFF_UP | IFF_RUNNING)) !=
229 		    (IFF_UP | IFF_RUNNING))
230 			senderr(ENETDOWN);
231 	}
232 #endif /* NCARP > 0 */
233 
234 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
235 		senderr(ENETDOWN);
236 
237 	switch (dst->sa_family) {
238 
239 #ifdef INET
240 	case AF_INET:
241 		if (m->m_flags & M_BCAST)
242 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
243 		else if (m->m_flags & M_MCAST)
244 			ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst);
245 		else if ((error = arpresolve(ifp, rt, m, dst, edst,
246 		    sizeof(edst))) != 0) {
247 			return error == EWOULDBLOCK ? 0 : error;
248 		}
249 		/* If broadcasting on a simplex interface, loopback a copy */
250 		if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
251 			mcopy = m_copy(m, 0, M_COPYALL);
252 		etype = htons(ETHERTYPE_IP);
253 		break;
254 
255 	case AF_ARP:
256 		ah = mtod(m, struct arphdr *);
257 		if (m->m_flags & M_BCAST)
258 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
259 		else {
260 			void *tha = ar_tha(ah);
261 
262 			if (tha == NULL) {
263 				/* fake with ARPHDR_IEEE1394 */
264 				m_freem(m);
265 				return 0;
266 			}
267 			memcpy(edst, tha, sizeof(edst));
268 		}
269 
270 		ah->ar_hrd = htons(ARPHRD_ETHER);
271 
272 		switch (ntohs(ah->ar_op)) {
273 		case ARPOP_REVREQUEST:
274 		case ARPOP_REVREPLY:
275 			etype = htons(ETHERTYPE_REVARP);
276 			break;
277 
278 		case ARPOP_REQUEST:
279 		case ARPOP_REPLY:
280 		default:
281 			etype = htons(ETHERTYPE_ARP);
282 		}
283 		break;
284 #endif
285 
286 #ifdef INET6
287 	case AF_INET6:
288 		if (m->m_flags & M_BCAST)
289 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
290 		else if (m->m_flags & M_MCAST) {
291 			ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr,
292 			    edst);
293 		} else {
294 			error = nd6_resolve(ifp, rt, m, dst, edst,
295 			    sizeof(edst));
296 			if (error != 0)
297 				return error == EWOULDBLOCK ? 0 : error;
298 		}
299 		etype = htons(ETHERTYPE_IPV6);
300 		break;
301 #endif
302 
303 #ifdef NETATALK
304 	case AF_APPLETALK: {
305 		struct ifaddr *ifa;
306 		int s;
307 
308 		KERNEL_LOCK(1, NULL);
309 
310 		if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) {
311 #ifdef NETATALKDEBUG
312 			printf("aarpresolv failed\n");
313 #endif
314 			KERNEL_UNLOCK_ONE(NULL);
315 			return (0);
316 		}
317 
318 		/*
319 		 * ifaddr is the first thing in at_ifaddr
320 		 */
321 		s = pserialize_read_enter();
322 		ifa = at_ifawithnet((const struct sockaddr_at *)dst, ifp);
323 		if (ifa == NULL) {
324 			pserialize_read_exit(s);
325 			KERNEL_UNLOCK_ONE(NULL);
326 			/* XXX error? */
327 			goto bad;
328 		}
329 		aa = (struct at_ifaddr *)ifa;
330 
331 		/*
332 		 * In the phase 2 case, we need to prepend an mbuf for the
333 		 * llc header.  Since we must preserve the value of m,
334 		 * which is passed to us by value, we m_copy() the first
335 		 * mbuf, and use it for our llc header.
336 		 */
337 		if (aa->aa_flags & AFA_PHASE2) {
338 			struct llc llc;
339 
340 			M_PREPEND(m, sizeof(struct llc), M_DONTWAIT);
341 			if (m == NULL) {
342 				KERNEL_UNLOCK_ONE(NULL);
343 				senderr(ENOBUFS);
344 			}
345 
346 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
347 			llc.llc_control = LLC_UI;
348 			memcpy(llc.llc_snap_org_code, at_org_code,
349 			    sizeof(llc.llc_snap_org_code));
350 			llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK);
351 			memcpy(mtod(m, void *), &llc, sizeof(struct llc));
352 		} else {
353 			etype = htons(ETHERTYPE_ATALK);
354 		}
355 		pserialize_read_exit(s);
356 		KERNEL_UNLOCK_ONE(NULL);
357 		break;
358 	}
359 #endif /* NETATALK */
360 
361 	case pseudo_AF_HDRCMPLT:
362 		hdrcmplt = 1;
363 		memcpy(esrc,
364 		    ((const struct ether_header *)dst->sa_data)->ether_shost,
365 		    sizeof(esrc));
366 		/* FALLTHROUGH */
367 
368 	case AF_UNSPEC:
369 		memcpy(edst,
370 		    ((const struct ether_header *)dst->sa_data)->ether_dhost,
371 		    sizeof(edst));
372 		/* AF_UNSPEC doesn't swap the byte order of the ether_type. */
373 		etype = ((const struct ether_header *)dst->sa_data)->ether_type;
374 		break;
375 
376 	default:
377 		printf("%s: can't handle af%d\n", ifp->if_xname,
378 			dst->sa_family);
379 		senderr(EAFNOSUPPORT);
380 	}
381 
382 #ifdef MPLS
383 	KERNEL_LOCK(1, NULL);
384 	{
385 		struct m_tag *mtag;
386 		mtag = m_tag_find(m, PACKET_TAG_MPLS, NULL);
387 		if (mtag != NULL) {
388 			/* Having the tag itself indicates it's MPLS */
389 			etype = htons(ETHERTYPE_MPLS);
390 			m_tag_delete(m, mtag);
391 		}
392 	}
393 	KERNEL_UNLOCK_ONE(NULL);
394 #endif
395 
396 	if (mcopy)
397 		(void)looutput(ifp, mcopy, dst, rt);
398 
399 	KASSERT((m->m_flags & M_PKTHDR) != 0);
400 
401 	/*
402 	 * If no ether type is set, this must be a 802.2 formatted packet.
403 	 */
404 	if (etype == 0)
405 		etype = htons(m->m_pkthdr.len);
406 
407 	/*
408 	 * Add local net header. If no space in first mbuf, allocate another.
409 	 */
410 	M_PREPEND(m, sizeof(struct ether_header), M_DONTWAIT);
411 	if (m == NULL)
412 		senderr(ENOBUFS);
413 
414 	eh = mtod(m, struct ether_header *);
415 	/* Note: etype is already in network byte order. */
416 	memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
417 	memcpy(eh->ether_dhost, edst, sizeof(edst));
418 	if (hdrcmplt)
419 		memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost));
420 	else
421 	 	memcpy(eh->ether_shost, CLLADDR(ifp->if_sadl),
422 		    sizeof(eh->ether_shost));
423 
424 #if NCARP > 0
425 	if (ifp0 != ifp && ifp0->if_type == IFT_CARP) {
426 	 	memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl),
427 		    sizeof(eh->ether_shost));
428 	}
429 #endif
430 
431 	if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0)
432 		return (error);
433 	if (m == NULL)
434 		return (0);
435 
436 #if NBRIDGE > 0
437 	/*
438 	 * Bridges require special output handling.
439 	 */
440 	if (ifp->if_bridge)
441 		return bridge_output(ifp, m, NULL, NULL);
442 #endif
443 
444 #if NCARP > 0
445 	if (ifp != ifp0)
446 		ifp0->if_obytes += m->m_pkthdr.len + ETHER_HDR_LEN;
447 #endif
448 
449 #ifdef ALTQ
450 	KERNEL_LOCK(1, NULL);
451 	/*
452 	 * If ALTQ is enabled on the parent interface, do
453 	 * classification; the queueing discipline might not
454 	 * require classification, but might require the
455 	 * address family/header pointer in the pktattr.
456 	 */
457 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
458 		altq_etherclassify(&ifp->if_snd, m);
459 	KERNEL_UNLOCK_ONE(NULL);
460 #endif
461 	return ifq_enqueue(ifp, m);
462 
463 bad:
464 	if (m)
465 		m_freem(m);
466 	return (error);
467 }
468 
469 #ifdef ALTQ
470 /*
471  * This routine is a slight hack to allow a packet to be classified
472  * if the Ethernet headers are present.  It will go away when ALTQ's
473  * classification engine understands link headers.
474  */
475 void
476 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m)
477 {
478 	struct ether_header *eh;
479 	struct mbuf *mtop = m;
480 	uint16_t ether_type;
481 	int hlen, af, hdrsize;
482 	void *hdr;
483 
484 	KASSERT((mtop->m_flags & M_PKTHDR) != 0);
485 
486 	hlen = ETHER_HDR_LEN;
487 	eh = mtod(m, struct ether_header *);
488 
489 	ether_type = htons(eh->ether_type);
490 
491 	if (ether_type < ETHERMTU) {
492 		/* LLC/SNAP */
493 		struct llc *llc = (struct llc *)(eh + 1);
494 		hlen += 8;
495 
496 		if (m->m_len < hlen ||
497 		    llc->llc_dsap != LLC_SNAP_LSAP ||
498 		    llc->llc_ssap != LLC_SNAP_LSAP ||
499 		    llc->llc_control != LLC_UI) {
500 			/* Not SNAP. */
501 			goto bad;
502 		}
503 
504 		ether_type = htons(llc->llc_un.type_snap.ether_type);
505 	}
506 
507 	switch (ether_type) {
508 	case ETHERTYPE_IP:
509 		af = AF_INET;
510 		hdrsize = 20;		/* sizeof(struct ip) */
511 		break;
512 
513 	case ETHERTYPE_IPV6:
514 		af = AF_INET6;
515 		hdrsize = 40;		/* sizeof(struct ip6_hdr) */
516 		break;
517 
518 	default:
519 		af = AF_UNSPEC;
520 		hdrsize = 0;
521 		break;
522 	}
523 
524 	while (m->m_len <= hlen) {
525 		hlen -= m->m_len;
526 		m = m->m_next;
527 		if (m == NULL)
528 			goto bad;
529 	}
530 
531 	if (m->m_len < (hlen + hdrsize)) {
532 		/*
533 		 * protocol header not in a single mbuf.
534 		 * We can't cope with this situation right
535 		 * now (but it shouldn't ever happen, really, anyhow).
536 		 */
537 #ifdef DEBUG
538 		printf("altq_etherclassify: headers span multiple mbufs: "
539 		    "%d < %d\n", m->m_len, (hlen + hdrsize));
540 #endif
541 		goto bad;
542 	}
543 
544 	m->m_data += hlen;
545 	m->m_len -= hlen;
546 
547 	hdr = mtod(m, void *);
548 
549 	if (ALTQ_NEEDS_CLASSIFY(ifq)) {
550 		mtop->m_pkthdr.pattr_class =
551 		    (*ifq->altq_classify)(ifq->altq_clfier, m, af);
552 	}
553 	mtop->m_pkthdr.pattr_af = af;
554 	mtop->m_pkthdr.pattr_hdr = hdr;
555 
556 	m->m_data -= hlen;
557 	m->m_len += hlen;
558 
559 	return;
560 
561 bad:
562 	mtop->m_pkthdr.pattr_class = NULL;
563 	mtop->m_pkthdr.pattr_hdr = NULL;
564 	mtop->m_pkthdr.pattr_af = AF_UNSPEC;
565 }
566 #endif /* ALTQ */
567 
568 /*
569  * Process a received Ethernet packet;
570  * the packet is in the mbuf chain m with
571  * the ether header.
572  */
573 void
574 ether_input(struct ifnet *ifp, struct mbuf *m)
575 {
576 	struct ethercom *ec = (struct ethercom *) ifp;
577 	pktqueue_t *pktq = NULL;
578 	struct ifqueue *inq = NULL;
579 	uint16_t etype;
580 	struct ether_header *eh;
581 	size_t ehlen;
582 	static int earlypkts;
583 	int isr = 0;
584 #if defined (LLC) || defined(NETATALK)
585 	struct llc *l;
586 #endif
587 
588 	KASSERT(!cpu_intr_p());
589 	KASSERT((m->m_flags & M_PKTHDR) != 0);
590 
591 	if ((ifp->if_flags & IFF_UP) == 0) {
592 		m_freem(m);
593 		return;
594 	}
595 
596 #ifdef MBUFTRACE
597 	m_claimm(m, &ec->ec_rx_mowner);
598 #endif
599 	eh = mtod(m, struct ether_header *);
600 	etype = ntohs(eh->ether_type);
601 	ehlen = sizeof(*eh);
602 
603 	if (__predict_false(earlypkts < 100 || !rnd_initial_entropy)) {
604 		rnd_add_data(NULL, eh, ehlen, 0);
605 		earlypkts++;
606 	}
607 
608 	/*
609 	 * Determine if the packet is within its size limits.
610 	 */
611 	if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len >
612 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
613 		mutex_enter(&bigpktpps_lock);
614 		if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count,
615 			    bigpktppslim)) {
616 			printf("%s: discarding oversize frame (len=%d)\n",
617 			    ifp->if_xname, m->m_pkthdr.len);
618 		}
619 		mutex_exit(&bigpktpps_lock);
620 		m_freem(m);
621 		return;
622 	}
623 
624 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
625 		/*
626 		 * If this is not a simplex interface, drop the packet
627 		 * if it came from us.
628 		 */
629 		if ((ifp->if_flags & IFF_SIMPLEX) == 0 &&
630 		    memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost,
631 		    ETHER_ADDR_LEN) == 0) {
632 			m_freem(m);
633 			return;
634 		}
635 
636 		if (memcmp(etherbroadcastaddr,
637 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
638 			m->m_flags |= M_BCAST;
639 		else
640 			m->m_flags |= M_MCAST;
641 		ifp->if_imcasts++;
642 	}
643 
644 	/* If the CRC is still on the packet, trim it off. */
645 	if (m->m_flags & M_HASFCS) {
646 		m_adj(m, -ETHER_CRC_LEN);
647 		m->m_flags &= ~M_HASFCS;
648 	}
649 
650 	ifp->if_ibytes += m->m_pkthdr.len;
651 
652 #if NCARP > 0
653 	if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) {
654 		/*
655 		 * clear M_PROMISC, in case the packets comes from a
656 		 * vlan
657 		 */
658 		m->m_flags &= ~M_PROMISC;
659 		if (carp_input(m, (uint8_t *)&eh->ether_shost,
660 		    (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0)
661 			return;
662 	}
663 #endif
664 
665 	if ((m->m_flags & (M_BCAST | M_MCAST | M_PROMISC)) == 0 &&
666 	    (ifp->if_flags & IFF_PROMISC) != 0 &&
667 	    memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost,
668 		   ETHER_ADDR_LEN) != 0) {
669 		m->m_flags |= M_PROMISC;
670 	}
671 
672 	if ((m->m_flags & M_PROMISC) == 0) {
673 		if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
674 			return;
675 		if (m == NULL)
676 			return;
677 
678 		eh = mtod(m, struct ether_header *);
679 		etype = ntohs(eh->ether_type);
680 	}
681 
682 #if NAGR > 0
683 	if (ifp->if_agrprivate &&
684 	    __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) {
685 		m->m_flags &= ~M_PROMISC;
686 		agr_input(ifp, m);
687 		return;
688 	}
689 #endif
690 
691 	/*
692 	 * If VLANs are configured on the interface, check to
693 	 * see if the device performed the decapsulation and
694 	 * provided us with the tag.
695 	 */
696 	if (ec->ec_nvlans && vlan_has_tag(m)) {
697 #if NVLAN > 0
698 		/*
699 		 * vlan_input() will either recursively call ether_input()
700 		 * or drop the packet.
701 		 */
702 		vlan_input(ifp, m);
703 #else
704 		m_freem(m);
705 #endif
706 		return;
707 	}
708 
709 	/*
710 	 * Handle protocols that expect to have the Ethernet header
711 	 * (and possibly FCS) intact.
712 	 */
713 	switch (etype) {
714 	case ETHERTYPE_VLAN: {
715 		struct ether_vlan_header *evl = (void *)eh;
716 
717 		/*
718 		 * If there is a tag of 0, then the VLAN header was probably
719 		 * just being used to store the priority.  Extract the ether
720 		 * type, and if IP or IPV6, let them deal with it.
721 		 */
722 		if (m->m_len >= sizeof(*evl) &&
723 		    EVL_VLANOFTAG(evl->evl_tag) == 0) {
724 			etype = ntohs(evl->evl_proto);
725 			ehlen = sizeof(*evl);
726 			if ((m->m_flags & M_PROMISC) == 0 &&
727 			    (etype == ETHERTYPE_IP ||
728 			     etype == ETHERTYPE_IPV6))
729 				break;
730 		}
731 
732 #if NVLAN > 0
733 		/*
734 		 * vlan_input() will either recursively call ether_input()
735 		 * or drop the packet.
736 		 */
737 		if (ec->ec_nvlans != 0)
738 			vlan_input(ifp, m);
739 		else
740 #endif
741 			m_freem(m);
742 
743 		return;
744 	}
745 
746 #if NPPPOE > 0
747 	case ETHERTYPE_PPPOEDISC:
748 		pppoedisc_input(ifp, m);
749 		return;
750 
751 	case ETHERTYPE_PPPOE:
752 		pppoe_input(ifp, m);
753 		return;
754 #endif
755 
756 	case ETHERTYPE_SLOWPROTOCOLS: {
757 		uint8_t subtype;
758 
759 		KASSERTMSG((m->m_pkthdr.len >= sizeof(*eh) + sizeof(subtype)),
760 			"too short slow protocol packet");
761 
762 		m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype);
763 		switch (subtype) {
764 #if NAGR > 0
765 		case SLOWPROTOCOLS_SUBTYPE_LACP:
766 			if (ifp->if_agrprivate) {
767 				ieee8023ad_lacp_input(ifp, m);
768 				return;
769 			}
770 			break;
771 
772 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
773 			if (ifp->if_agrprivate) {
774 				ieee8023ad_marker_input(ifp, m);
775 				return;
776 			}
777 			break;
778 #endif /* NAGR > 0 */
779 
780 		default:
781 			if (subtype == 0 || subtype > 10) {
782 				/* illegal value */
783 				m_freem(m);
784 				return;
785 			}
786 			/* unknown subtype */
787 			break;
788 		}
789 		/* FALLTHROUGH */
790 	}
791 
792 	default:
793 		if (m->m_flags & M_PROMISC) {
794 			m_freem(m);
795 			return;
796 		}
797 	}
798 
799 	/* If the CRC is still on the packet, trim it off. */
800 	if (m->m_flags & M_HASFCS) {
801 		m_adj(m, -ETHER_CRC_LEN);
802 		m->m_flags &= ~M_HASFCS;
803 	}
804 
805 	if (etype > ETHERMTU + sizeof(struct ether_header)) {
806 		/* Strip off the Ethernet header. */
807 		m_adj(m, ehlen);
808 
809 		switch (etype) {
810 #ifdef INET
811 		case ETHERTYPE_IP:
812 #ifdef GATEWAY
813 			if (ipflow_fastforward(m))
814 				return;
815 #endif
816 			pktq = ip_pktq;
817 			break;
818 
819 		case ETHERTYPE_ARP:
820 			isr = NETISR_ARP;
821 			inq = &arpintrq;
822 			break;
823 
824 		case ETHERTYPE_REVARP:
825 			revarpinput(m);	/* XXX queue? */
826 			return;
827 #endif
828 
829 #ifdef INET6
830 		case ETHERTYPE_IPV6:
831 			if (__predict_false(!in6_present)) {
832 				m_freem(m);
833 				return;
834 			}
835 #ifdef GATEWAY
836 			if (ip6flow_fastforward(&m))
837 				return;
838 #endif
839 			pktq = ip6_pktq;
840 			break;
841 #endif
842 
843 #ifdef NETATALK
844 		case ETHERTYPE_ATALK:
845 			isr = NETISR_ATALK;
846 			inq = &atintrq1;
847 			break;
848 
849 		case ETHERTYPE_AARP:
850 			aarpinput(ifp, m); /* XXX queue? */
851 			return;
852 #endif
853 
854 #ifdef MPLS
855 		case ETHERTYPE_MPLS:
856 			isr = NETISR_MPLS;
857 			inq = &mplsintrq;
858 			break;
859 #endif
860 
861 		default:
862 			m_freem(m);
863 			return;
864 		}
865 	} else {
866 		KASSERT(ehlen == sizeof(*eh));
867 #if defined (LLC) || defined (NETATALK)
868 		if (m->m_len < sizeof(*eh) + sizeof(struct llc)) {
869 			goto dropanyway;
870 		}
871 		l = (struct llc *)(eh+1);
872 
873 		switch (l->llc_dsap) {
874 #ifdef NETATALK
875 		case LLC_SNAP_LSAP:
876 			switch (l->llc_control) {
877 			case LLC_UI:
878 				if (l->llc_ssap != LLC_SNAP_LSAP) {
879 					goto dropanyway;
880 				}
881 
882 				if (memcmp(&(l->llc_snap_org_code)[0],
883 				    at_org_code, sizeof(at_org_code)) == 0 &&
884 				    ntohs(l->llc_snap_ether_type) ==
885 				    ETHERTYPE_ATALK) {
886 					inq = &atintrq2;
887 					m_adj(m, sizeof(struct ether_header)
888 					    + sizeof(struct llc));
889 					isr = NETISR_ATALK;
890 					break;
891 				}
892 
893 				if (memcmp(&(l->llc_snap_org_code)[0],
894 				    aarp_org_code,
895 				    sizeof(aarp_org_code)) == 0 &&
896 				    ntohs(l->llc_snap_ether_type) ==
897 				    ETHERTYPE_AARP) {
898 					m_adj(m, sizeof(struct ether_header)
899 					    + sizeof(struct llc));
900 					aarpinput(ifp, m); /* XXX */
901 					return;
902 				}
903 
904 			default:
905 				goto dropanyway;
906 			}
907 			break;
908 #endif
909 		dropanyway:
910 		default:
911 			m_freem(m);
912 			return;
913 		}
914 #else /* LLC || NETATALK */
915 		m_freem(m);
916 		return;
917 #endif /* LLC || NETATALK */
918 	}
919 
920 	if (__predict_true(pktq)) {
921 #ifdef NET_MPSAFE
922 		const u_int h = curcpu()->ci_index;
923 #else
924 		const uint32_t h = pktq_rps_hash(m);
925 #endif
926 		if (__predict_false(!pktq_enqueue(pktq, m, h))) {
927 			m_freem(m);
928 		}
929 		return;
930 	}
931 
932 	if (__predict_false(!inq)) {
933 		/* Should not happen. */
934 		m_freem(m);
935 		return;
936 	}
937 
938 	IFQ_LOCK(inq);
939 	if (IF_QFULL(inq)) {
940 		IF_DROP(inq);
941 		IFQ_UNLOCK(inq);
942 		m_freem(m);
943 	} else {
944 		IF_ENQUEUE(inq, m);
945 		IFQ_UNLOCK(inq);
946 		schednetisr(isr);
947 	}
948 }
949 
950 /*
951  * Convert Ethernet address to printable (loggable) representation.
952  */
953 char *
954 ether_sprintf(const u_char *ap)
955 {
956 	static char etherbuf[3 * ETHER_ADDR_LEN];
957 	return ether_snprintf(etherbuf, sizeof(etherbuf), ap);
958 }
959 
960 char *
961 ether_snprintf(char *buf, size_t len, const u_char *ap)
962 {
963 	char *cp = buf;
964 	size_t i;
965 
966 	for (i = 0; i < len / 3; i++) {
967 		*cp++ = hexdigits[*ap >> 4];
968 		*cp++ = hexdigits[*ap++ & 0xf];
969 		*cp++ = ':';
970 	}
971 	*--cp = '\0';
972 	return buf;
973 }
974 
975 /*
976  * Perform common duties while attaching to interface list
977  */
978 void
979 ether_ifattach(struct ifnet *ifp, const uint8_t *lla)
980 {
981 	struct ethercom *ec = (struct ethercom *)ifp;
982 
983 	ifp->if_type = IFT_ETHER;
984 	ifp->if_hdrlen = ETHER_HDR_LEN;
985 	ifp->if_dlt = DLT_EN10MB;
986 	ifp->if_mtu = ETHERMTU;
987 	ifp->if_output = ether_output;
988 	ifp->_if_input = ether_input;
989 	if (ifp->if_baudrate == 0)
990 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
991 
992 	if (lla != NULL)
993 		if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla));
994 
995 	LIST_INIT(&ec->ec_multiaddrs);
996 	ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
997 	ifp->if_broadcastaddr = etherbroadcastaddr;
998 	bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header));
999 #ifdef MBUFTRACE
1000 	strlcpy(ec->ec_tx_mowner.mo_name, ifp->if_xname,
1001 	    sizeof(ec->ec_tx_mowner.mo_name));
1002 	strlcpy(ec->ec_tx_mowner.mo_descr, "tx",
1003 	    sizeof(ec->ec_tx_mowner.mo_descr));
1004 	strlcpy(ec->ec_rx_mowner.mo_name, ifp->if_xname,
1005 	    sizeof(ec->ec_rx_mowner.mo_name));
1006 	strlcpy(ec->ec_rx_mowner.mo_descr, "rx",
1007 	    sizeof(ec->ec_rx_mowner.mo_descr));
1008 	MOWNER_ATTACH(&ec->ec_tx_mowner);
1009 	MOWNER_ATTACH(&ec->ec_rx_mowner);
1010 	ifp->if_mowner = &ec->ec_tx_mowner;
1011 #endif
1012 }
1013 
1014 void
1015 ether_ifdetach(struct ifnet *ifp)
1016 {
1017 	struct ethercom *ec = (void *) ifp;
1018 	struct ether_multi *enm;
1019 
1020 	/*
1021 	 * Prevent further calls to ioctl (for example turning off
1022 	 * promiscuous mode from the bridge code), which eventually can
1023 	 * call if_init() which can cause panics because the interface
1024 	 * is in the process of being detached. Return device not configured
1025 	 * instead.
1026 	 */
1027 	ifp->if_ioctl = (int (*)(struct ifnet *, u_long, void *))enxio;
1028 
1029 #if NBRIDGE > 0
1030 	if (ifp->if_bridge)
1031 		bridge_ifdetach(ifp);
1032 #endif
1033 	bpf_detach(ifp);
1034 #if NVLAN > 0
1035 	if (ec->ec_nvlans)
1036 		vlan_ifdetach(ifp);
1037 #endif
1038 
1039 	ETHER_LOCK(ec);
1040 	while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) {
1041 		LIST_REMOVE(enm, enm_list);
1042 		kmem_free(enm, sizeof(*enm));
1043 		ec->ec_multicnt--;
1044 	}
1045 	ETHER_UNLOCK(ec);
1046 
1047 	mutex_obj_free(ec->ec_lock);
1048 	ec->ec_lock = NULL;
1049 
1050 	ifp->if_mowner = NULL;
1051 	MOWNER_DETACH(&ec->ec_rx_mowner);
1052 	MOWNER_DETACH(&ec->ec_tx_mowner);
1053 }
1054 
1055 #if 0
1056 /*
1057  * This is for reference.  We have a table-driven version
1058  * of the little-endian crc32 generator, which is faster
1059  * than the double-loop.
1060  */
1061 uint32_t
1062 ether_crc32_le(const uint8_t *buf, size_t len)
1063 {
1064 	uint32_t c, crc, carry;
1065 	size_t i, j;
1066 
1067 	crc = 0xffffffffU;	/* initial value */
1068 
1069 	for (i = 0; i < len; i++) {
1070 		c = buf[i];
1071 		for (j = 0; j < 8; j++) {
1072 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1073 			crc >>= 1;
1074 			c >>= 1;
1075 			if (carry)
1076 				crc = (crc ^ ETHER_CRC_POLY_LE);
1077 		}
1078 	}
1079 
1080 	return (crc);
1081 }
1082 #else
1083 uint32_t
1084 ether_crc32_le(const uint8_t *buf, size_t len)
1085 {
1086 	static const uint32_t crctab[] = {
1087 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1088 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1089 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1090 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1091 	};
1092 	uint32_t crc;
1093 	size_t i;
1094 
1095 	crc = 0xffffffffU;	/* initial value */
1096 
1097 	for (i = 0; i < len; i++) {
1098 		crc ^= buf[i];
1099 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1100 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1101 	}
1102 
1103 	return (crc);
1104 }
1105 #endif
1106 
1107 uint32_t
1108 ether_crc32_be(const uint8_t *buf, size_t len)
1109 {
1110 	uint32_t c, crc, carry;
1111 	size_t i, j;
1112 
1113 	crc = 0xffffffffU;	/* initial value */
1114 
1115 	for (i = 0; i < len; i++) {
1116 		c = buf[i];
1117 		for (j = 0; j < 8; j++) {
1118 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1119 			crc <<= 1;
1120 			c >>= 1;
1121 			if (carry)
1122 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1123 		}
1124 	}
1125 
1126 	return (crc);
1127 }
1128 
1129 #ifdef INET
1130 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] =
1131     { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 };
1132 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] =
1133     { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff };
1134 #endif
1135 #ifdef INET6
1136 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] =
1137     { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 };
1138 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] =
1139     { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff };
1140 #endif
1141 
1142 /*
1143  * ether_aton implementation, not using a static buffer.
1144  */
1145 int
1146 ether_aton_r(u_char *dest, size_t len, const char *str)
1147 {
1148 	const u_char *cp = (const void *)str;
1149 	u_char *ep;
1150 
1151 #define atox(c)	(((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10))
1152 
1153 	if (len < ETHER_ADDR_LEN)
1154 		return ENOSPC;
1155 
1156 	ep = dest + ETHER_ADDR_LEN;
1157 
1158 	while (*cp) {
1159 		if (!isxdigit(*cp))
1160 			return EINVAL;
1161 
1162 		*dest = atox(*cp);
1163 		cp++;
1164 		if (isxdigit(*cp)) {
1165 			*dest = (*dest << 4) | atox(*cp);
1166 			cp++;
1167 		}
1168 		dest++;
1169 
1170 		if (dest == ep)
1171 			return (*cp == '\0') ? 0 : ENAMETOOLONG;
1172 
1173 		switch (*cp) {
1174 		case ':':
1175 		case '-':
1176 		case '.':
1177 			cp++;
1178 			break;
1179 		}
1180 	}
1181 	return ENOBUFS;
1182 }
1183 
1184 /*
1185  * Convert a sockaddr into an Ethernet address or range of Ethernet
1186  * addresses.
1187  */
1188 int
1189 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN],
1190     uint8_t addrhi[ETHER_ADDR_LEN])
1191 {
1192 #ifdef INET
1193 	const struct sockaddr_in *sin;
1194 #endif /* INET */
1195 #ifdef INET6
1196 	const struct sockaddr_in6 *sin6;
1197 #endif /* INET6 */
1198 
1199 	switch (sa->sa_family) {
1200 
1201 	case AF_UNSPEC:
1202 		memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN);
1203 		memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1204 		break;
1205 
1206 #ifdef INET
1207 	case AF_INET:
1208 		sin = satocsin(sa);
1209 		if (sin->sin_addr.s_addr == INADDR_ANY) {
1210 			/*
1211 			 * An IP address of INADDR_ANY means listen to
1212 			 * or stop listening to all of the Ethernet
1213 			 * multicast addresses used for IP.
1214 			 * (This is for the sake of IP multicast routers.)
1215 			 */
1216 			memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN);
1217 			memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN);
1218 		} else {
1219 			ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo);
1220 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1221 		}
1222 		break;
1223 #endif
1224 #ifdef INET6
1225 	case AF_INET6:
1226 		sin6 = satocsin6(sa);
1227 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1228 			/*
1229 			 * An IP6 address of 0 means listen to or stop
1230 			 * listening to all of the Ethernet multicast
1231 			 * address used for IP6.
1232 			 * (This is used for multicast routers.)
1233 			 */
1234 			memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN);
1235 			memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN);
1236 		} else {
1237 			ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo);
1238 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1239 		}
1240 		break;
1241 #endif
1242 
1243 	default:
1244 		return EAFNOSUPPORT;
1245 	}
1246 	return 0;
1247 }
1248 
1249 /*
1250  * Add an Ethernet multicast address or range of addresses to the list for a
1251  * given interface.
1252  */
1253 int
1254 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec)
1255 {
1256 	struct ether_multi *enm, *_enm;
1257 	u_char addrlo[ETHER_ADDR_LEN];
1258 	u_char addrhi[ETHER_ADDR_LEN];
1259 	int error = 0;
1260 
1261 	/* Allocate out of lock */
1262 	enm = kmem_alloc(sizeof(*enm), KM_SLEEP);
1263 
1264 	ETHER_LOCK(ec);
1265 	error = ether_multiaddr(sa, addrlo, addrhi);
1266 	if (error != 0)
1267 		goto out;
1268 
1269 	/*
1270 	 * Verify that we have valid Ethernet multicast addresses.
1271 	 */
1272 	if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) {
1273 		error = EINVAL;
1274 		goto out;
1275 	}
1276 
1277 	/*
1278 	 * See if the address range is already in the list.
1279 	 */
1280 	ETHER_LOOKUP_MULTI(addrlo, addrhi, ec, _enm);
1281 	if (_enm != NULL) {
1282 		/*
1283 		 * Found it; just increment the reference count.
1284 		 */
1285 		++_enm->enm_refcount;
1286 		error = 0;
1287 		goto out;
1288 	}
1289 
1290 	/*
1291 	 * Link a new multicast record into the interface's multicast list.
1292 	 */
1293 	memcpy(enm->enm_addrlo, addrlo, ETHER_ADDR_LEN);
1294 	memcpy(enm->enm_addrhi, addrhi, ETHER_ADDR_LEN);
1295 	enm->enm_refcount = 1;
1296 	LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list);
1297 	ec->ec_multicnt++;
1298 
1299 	/*
1300 	 * Return ENETRESET to inform the driver that the list has changed
1301 	 * and its reception filter should be adjusted accordingly.
1302 	 */
1303 	error = ENETRESET;
1304 	enm = NULL;
1305 
1306 out:
1307 	ETHER_UNLOCK(ec);
1308 	if (enm != NULL)
1309 		kmem_free(enm, sizeof(*enm));
1310 	return error;
1311 }
1312 
1313 /*
1314  * Delete a multicast address record.
1315  */
1316 int
1317 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec)
1318 {
1319 	struct ether_multi *enm;
1320 	u_char addrlo[ETHER_ADDR_LEN];
1321 	u_char addrhi[ETHER_ADDR_LEN];
1322 	int error;
1323 
1324 	ETHER_LOCK(ec);
1325 	error = ether_multiaddr(sa, addrlo, addrhi);
1326 	if (error != 0)
1327 		goto error;
1328 
1329 	/*
1330 	 * Look up the address in our list.
1331 	 */
1332 	ETHER_LOOKUP_MULTI(addrlo, addrhi, ec, enm);
1333 	if (enm == NULL) {
1334 		error = ENXIO;
1335 		goto error;
1336 	}
1337 	if (--enm->enm_refcount != 0) {
1338 		/*
1339 		 * Still some claims to this record.
1340 		 */
1341 		error = 0;
1342 		goto error;
1343 	}
1344 
1345 	/*
1346 	 * No remaining claims to this record; unlink and free it.
1347 	 */
1348 	LIST_REMOVE(enm, enm_list);
1349 	ec->ec_multicnt--;
1350 	ETHER_UNLOCK(ec);
1351 	kmem_free(enm, sizeof(*enm));
1352 
1353 	/*
1354 	 * Return ENETRESET to inform the driver that the list has changed
1355 	 * and its reception filter should be adjusted accordingly.
1356 	 */
1357 	return ENETRESET;
1358 
1359 error:
1360 	ETHER_UNLOCK(ec);
1361 	return error;
1362 }
1363 
1364 void
1365 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb)
1366 {
1367 	ec->ec_ifflags_cb = cb;
1368 }
1369 
1370 /*
1371  * Common ioctls for Ethernet interfaces.  Note, we must be
1372  * called at splnet().
1373  */
1374 int
1375 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data)
1376 {
1377 	struct ethercom *ec = (void *) ifp;
1378 	struct eccapreq *eccr;
1379 	struct ifreq *ifr = (struct ifreq *)data;
1380 	struct if_laddrreq *iflr = data;
1381 	const struct sockaddr_dl *sdl;
1382 	static const uint8_t zero[ETHER_ADDR_LEN];
1383 	int error;
1384 
1385 	switch (cmd) {
1386 	case SIOCINITIFADDR:
1387 	    {
1388 		struct ifaddr *ifa = (struct ifaddr *)data;
1389 		if (ifa->ifa_addr->sa_family != AF_LINK
1390 		    && (ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
1391 		       (IFF_UP | IFF_RUNNING)) {
1392 			ifp->if_flags |= IFF_UP;
1393 			if ((error = (*ifp->if_init)(ifp)) != 0)
1394 				return error;
1395 		}
1396 #ifdef INET
1397 		if (ifa->ifa_addr->sa_family == AF_INET)
1398 			arp_ifinit(ifp, ifa);
1399 #endif
1400 		return 0;
1401 	    }
1402 
1403 	case SIOCSIFMTU:
1404 	    {
1405 		int maxmtu;
1406 
1407 		if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU)
1408 			maxmtu = ETHERMTU_JUMBO;
1409 		else
1410 			maxmtu = ETHERMTU;
1411 
1412 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu)
1413 			return EINVAL;
1414 		else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET)
1415 			return error;
1416 		else if (ifp->if_flags & IFF_UP) {
1417 			/* Make sure the device notices the MTU change. */
1418 			return (*ifp->if_init)(ifp);
1419 		} else
1420 			return 0;
1421 	    }
1422 
1423 	case SIOCSIFFLAGS:
1424 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
1425 			return error;
1426 		switch (ifp->if_flags & (IFF_UP | IFF_RUNNING)) {
1427 		case IFF_RUNNING:
1428 			/*
1429 			 * If interface is marked down and it is running,
1430 			 * then stop and disable it.
1431 			 */
1432 			(*ifp->if_stop)(ifp, 1);
1433 			break;
1434 		case IFF_UP:
1435 			/*
1436 			 * If interface is marked up and it is stopped, then
1437 			 * start it.
1438 			 */
1439 			return (*ifp->if_init)(ifp);
1440 		case IFF_UP | IFF_RUNNING:
1441 			error = 0;
1442 			if (ec->ec_ifflags_cb != NULL) {
1443 				error = (*ec->ec_ifflags_cb)(ec);
1444 				if (error == ENETRESET) {
1445 					/*
1446 					 * Reset the interface to pick up
1447 					 * changes in any other flags that
1448 					 * affect the hardware state.
1449 					 */
1450 					return (*ifp->if_init)(ifp);
1451 				}
1452 			} else
1453 				error = (*ifp->if_init)(ifp);
1454 			return error;
1455 		case 0:
1456 			break;
1457 		}
1458 		return 0;
1459 	case SIOCGETHERCAP:
1460 		eccr = (struct eccapreq *)data;
1461 		eccr->eccr_capabilities = ec->ec_capabilities;
1462 		eccr->eccr_capenable = ec->ec_capenable;
1463 		return 0;
1464 	case SIOCADDMULTI:
1465 		return ether_addmulti(ifreq_getaddr(cmd, ifr), ec);
1466 	case SIOCDELMULTI:
1467 		return ether_delmulti(ifreq_getaddr(cmd, ifr), ec);
1468 	case SIOCSIFMEDIA:
1469 	case SIOCGIFMEDIA:
1470 		if (ec->ec_mii == NULL)
1471 			return ENOTTY;
1472 		return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media, cmd);
1473 	case SIOCALIFADDR:
1474 		sdl = satocsdl(sstocsa(&iflr->addr));
1475 		if (sdl->sdl_family != AF_LINK)
1476 			;
1477 		else if (ETHER_IS_MULTICAST(CLLADDR(sdl)))
1478 			return EINVAL;
1479 		else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0)
1480 			return EINVAL;
1481 		/*FALLTHROUGH*/
1482 	default:
1483 		return ifioctl_common(ifp, cmd, data);
1484 	}
1485 	return 0;
1486 }
1487 
1488 /*
1489  * Enable/disable passing VLAN packets if the parent interface supports it.
1490  * Return:
1491  * 	 0: Ok
1492  *	-1: Parent interface does not support vlans
1493  *	>0: Error
1494  */
1495 int
1496 ether_enable_vlan_mtu(struct ifnet *ifp)
1497 {
1498 	int error;
1499 	struct ethercom *ec = (void *)ifp;
1500 
1501 	/* Parent does not support VLAN's */
1502 	if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0)
1503 		return -1;
1504 
1505 	/*
1506 	 * Parent supports the VLAN_MTU capability,
1507 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames;
1508 	 * enable it.
1509 	 */
1510 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1511 
1512 	/* Interface is down, defer for later */
1513 	if ((ifp->if_flags & IFF_UP) == 0)
1514 		return 0;
1515 
1516 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1517 		return 0;
1518 
1519 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1520 	return error;
1521 }
1522 
1523 int
1524 ether_disable_vlan_mtu(struct ifnet *ifp)
1525 {
1526 	int error;
1527 	struct ethercom *ec = (void *)ifp;
1528 
1529 	/* We still have VLAN's, defer for later */
1530 	if (ec->ec_nvlans != 0)
1531 		return 0;
1532 
1533 	/* Parent does not support VLAB's, nothing to do. */
1534 	if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0)
1535 		return -1;
1536 
1537 	/*
1538 	 * Disable Tx/Rx of VLAN-sized frames.
1539 	 */
1540 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1541 
1542 	/* Interface is down, defer for later */
1543 	if ((ifp->if_flags & IFF_UP) == 0)
1544 		return 0;
1545 
1546 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1547 		return 0;
1548 
1549 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1550 	return error;
1551 }
1552 
1553 static int
1554 ether_multicast_sysctl(SYSCTLFN_ARGS)
1555 {
1556 	struct ether_multi *enm;
1557 	struct ifnet *ifp;
1558 	struct ethercom *ec;
1559 	int error = 0;
1560 	size_t written;
1561 	struct psref psref;
1562 	int bound;
1563 	unsigned int multicnt;
1564 	struct ether_multi_sysctl *addrs;
1565 	int i;
1566 
1567 	if (namelen != 1)
1568 		return EINVAL;
1569 
1570 	bound = curlwp_bind();
1571 	ifp = if_get_byindex(name[0], &psref);
1572 	if (ifp == NULL) {
1573 		error = ENODEV;
1574 		goto out;
1575 	}
1576 	if (ifp->if_type != IFT_ETHER) {
1577 		if_put(ifp, &psref);
1578 		*oldlenp = 0;
1579 		goto out;
1580 	}
1581 	ec = (struct ethercom *)ifp;
1582 
1583 	if (oldp == NULL) {
1584 		if_put(ifp, &psref);
1585 		*oldlenp = ec->ec_multicnt * sizeof(*addrs);
1586 		goto out;
1587 	}
1588 
1589 	/*
1590 	 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which
1591 	 * is sleepable, while holding it. Copy data to a local buffer first
1592 	 * with the lock taken and then call sysctl_copyout without holding it.
1593 	 */
1594 retry:
1595 	multicnt = ec->ec_multicnt;
1596 
1597 	if (multicnt == 0) {
1598 		if_put(ifp, &psref);
1599 		*oldlenp = 0;
1600 		goto out;
1601 	}
1602 
1603 	addrs = kmem_zalloc(sizeof(*addrs) * multicnt, KM_SLEEP);
1604 
1605 	ETHER_LOCK(ec);
1606 	if (multicnt != ec->ec_multicnt) {
1607 		/* The number of multicast addresses has changed */
1608 		ETHER_UNLOCK(ec);
1609 		kmem_free(addrs, sizeof(*addrs) * multicnt);
1610 		goto retry;
1611 	}
1612 
1613 	i = 0;
1614 	LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) {
1615 		struct ether_multi_sysctl *addr = &addrs[i];
1616 		addr->enm_refcount = enm->enm_refcount;
1617 		memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN);
1618 		memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN);
1619 		i++;
1620 	}
1621 	ETHER_UNLOCK(ec);
1622 
1623 	error = 0;
1624 	written = 0;
1625 	for (i = 0; i < multicnt; i++) {
1626 		struct ether_multi_sysctl *addr = &addrs[i];
1627 
1628 		if (written + sizeof(*addr) > *oldlenp)
1629 			break;
1630 		error = sysctl_copyout(l, addr, oldp, sizeof(*addr));
1631 		if (error)
1632 			break;
1633 		written += sizeof(*addr);
1634 		oldp = (char *)oldp + sizeof(*addr);
1635 	}
1636 	kmem_free(addrs, sizeof(*addrs) * multicnt);
1637 
1638 	if_put(ifp, &psref);
1639 
1640 	*oldlenp = written;
1641 out:
1642 	curlwp_bindx(bound);
1643 	return error;
1644 }
1645 
1646 static void
1647 ether_sysctl_setup(struct sysctllog **clog)
1648 {
1649 	const struct sysctlnode *rnode = NULL;
1650 
1651 	sysctl_createv(clog, 0, NULL, &rnode,
1652 		       CTLFLAG_PERMANENT,
1653 		       CTLTYPE_NODE, "ether",
1654 		       SYSCTL_DESCR("Ethernet-specific information"),
1655 		       NULL, 0, NULL, 0,
1656 		       CTL_NET, CTL_CREATE, CTL_EOL);
1657 
1658 	sysctl_createv(clog, 0, &rnode, NULL,
1659 		       CTLFLAG_PERMANENT,
1660 		       CTLTYPE_NODE, "multicast",
1661 		       SYSCTL_DESCR("multicast addresses"),
1662 		       ether_multicast_sysctl, 0, NULL, 0,
1663 		       CTL_CREATE, CTL_EOL);
1664 }
1665 
1666 void
1667 etherinit(void)
1668 {
1669 
1670 	mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET);
1671 	ether_sysctl_setup(NULL);
1672 }
1673