xref: /netbsd-src/sys/net/if_ethersubr.c (revision 42b9e898991e23b560315a9b1da6a36a39d4351b)
1 /*	$NetBSD: if_ethersubr.c,v 1.284 2020/04/30 03:29:55 riastradh 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 1.284 2020/04/30 03:29:55 riastradh 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/entropy.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 #include "carp.h"
146 #if NCARP > 0
147 #include <netinet/ip_carp.h>
148 #endif
149 
150 #ifdef NETATALK
151 #include <netatalk/at.h>
152 #include <netatalk/at_var.h>
153 #include <netatalk/at_extern.h>
154 
155 #define llc_snap_org_code llc_un.type_snap.org_code
156 #define llc_snap_ether_type llc_un.type_snap.ether_type
157 
158 extern u_char	at_org_code[3];
159 extern u_char	aarp_org_code[3];
160 #endif /* NETATALK */
161 
162 #ifdef MPLS
163 #include <netmpls/mpls.h>
164 #include <netmpls/mpls_var.h>
165 #endif
166 
167 #ifdef DIAGNOSTIC
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 #endif
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
202 #ifdef NETATALK
203 	struct at_ifaddr *aa;
204 #endif
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
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 {
246 			error = arpresolve(ifp0, rt, m, dst, edst, sizeof(edst));
247 			if (error)
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_copypacket(m, M_DONTWAIT);
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 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
260 		} else {
261 			void *tha = ar_tha(ah);
262 
263 			if (tha == NULL) {
264 				/* fake with ARPHRD_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 		break;
285 #endif
286 
287 #ifdef INET6
288 	case AF_INET6:
289 		if (m->m_flags & M_BCAST) {
290 			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(ifp0, rt, m, dst, edst,
296 			    sizeof(edst));
297 			if (error)
298 				return (error == EWOULDBLOCK) ? 0 : error;
299 		}
300 		etype = htons(ETHERTYPE_IPV6);
301 		break;
302 #endif
303 
304 #ifdef NETATALK
305 	case AF_APPLETALK: {
306 		struct ifaddr *ifa;
307 		int s;
308 
309 		KERNEL_LOCK(1, NULL);
310 
311 		if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) {
312 			KERNEL_UNLOCK_ONE(NULL);
313 			return 0;
314 		}
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 			senderr(EADDRNOTAVAIL);
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.
331 		 */
332 		if (aa->aa_flags & AFA_PHASE2) {
333 			struct llc llc;
334 
335 			M_PREPEND(m, sizeof(struct llc), M_DONTWAIT);
336 			if (m == NULL) {
337 				pserialize_read_exit(s);
338 				KERNEL_UNLOCK_ONE(NULL);
339 				senderr(ENOBUFS);
340 			}
341 
342 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
343 			llc.llc_control = LLC_UI;
344 			memcpy(llc.llc_snap_org_code, at_org_code,
345 			    sizeof(llc.llc_snap_org_code));
346 			llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK);
347 			memcpy(mtod(m, void *), &llc, sizeof(struct llc));
348 		} else {
349 			etype = htons(ETHERTYPE_ATALK);
350 		}
351 		pserialize_read_exit(s);
352 		KERNEL_UNLOCK_ONE(NULL);
353 		break;
354 	}
355 #endif /* NETATALK */
356 
357 	case pseudo_AF_HDRCMPLT:
358 		hdrcmplt = 1;
359 		memcpy(esrc,
360 		    ((const struct ether_header *)dst->sa_data)->ether_shost,
361 		    sizeof(esrc));
362 		/* FALLTHROUGH */
363 
364 	case AF_UNSPEC:
365 		memcpy(edst,
366 		    ((const struct ether_header *)dst->sa_data)->ether_dhost,
367 		    sizeof(edst));
368 		/* AF_UNSPEC doesn't swap the byte order of the ether_type. */
369 		etype = ((const struct ether_header *)dst->sa_data)->ether_type;
370 		break;
371 
372 	default:
373 		printf("%s: can't handle af%d\n", ifp->if_xname,
374 		    dst->sa_family);
375 		senderr(EAFNOSUPPORT);
376 	}
377 
378 #ifdef MPLS
379 	{
380 		struct m_tag *mtag;
381 		mtag = m_tag_find(m, PACKET_TAG_MPLS);
382 		if (mtag != NULL) {
383 			/* Having the tag itself indicates it's MPLS */
384 			etype = htons(ETHERTYPE_MPLS);
385 			m_tag_delete(m, mtag);
386 		}
387 	}
388 #endif
389 
390 	if (mcopy)
391 		(void)looutput(ifp, mcopy, dst, rt);
392 
393 	KASSERT((m->m_flags & M_PKTHDR) != 0);
394 
395 	/*
396 	 * If no ether type is set, this must be a 802.2 formatted packet.
397 	 */
398 	if (etype == 0)
399 		etype = htons(m->m_pkthdr.len);
400 
401 	/*
402 	 * Add local net header. If no space in first mbuf, allocate another.
403 	 */
404 	M_PREPEND(m, sizeof(struct ether_header), M_DONTWAIT);
405 	if (m == NULL)
406 		senderr(ENOBUFS);
407 
408 	eh = mtod(m, struct ether_header *);
409 	/* Note: etype is already in network byte order. */
410 	memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
411 	memcpy(eh->ether_dhost, edst, sizeof(edst));
412 	if (hdrcmplt) {
413 		memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost));
414 	} else {
415 	 	memcpy(eh->ether_shost, CLLADDR(ifp->if_sadl),
416 		    sizeof(eh->ether_shost));
417 	}
418 
419 #if NCARP > 0
420 	if (ifp0 != ifp && ifp0->if_type == IFT_CARP) {
421 	 	memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl),
422 		    sizeof(eh->ether_shost));
423 	}
424 #endif
425 
426 	if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0)
427 		return error;
428 	if (m == NULL)
429 		return 0;
430 
431 #if NBRIDGE > 0
432 	/*
433 	 * Bridges require special output handling.
434 	 */
435 	if (ifp->if_bridge)
436 		return bridge_output(ifp, m, NULL, NULL);
437 #endif
438 
439 #if NCARP > 0
440 	if (ifp != ifp0)
441 		if_statadd(ifp0, if_obytes, m->m_pkthdr.len + ETHER_HDR_LEN);
442 #endif
443 
444 #ifdef ALTQ
445 	KERNEL_LOCK(1, NULL);
446 	/*
447 	 * If ALTQ is enabled on the parent interface, do
448 	 * classification; the queueing discipline might not
449 	 * require classification, but might require the
450 	 * address family/header pointer in the pktattr.
451 	 */
452 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
453 		altq_etherclassify(&ifp->if_snd, m);
454 	KERNEL_UNLOCK_ONE(NULL);
455 #endif
456 	return ifq_enqueue(ifp, m);
457 
458 bad:
459 	if (m)
460 		m_freem(m);
461 	return error;
462 }
463 
464 #ifdef ALTQ
465 /*
466  * This routine is a slight hack to allow a packet to be classified
467  * if the Ethernet headers are present.  It will go away when ALTQ's
468  * classification engine understands link headers.
469  *
470  * XXX: We may need to do m_pullups here. First to ensure struct ether_header
471  * is indeed contiguous, then to read the LLC and so on.
472  */
473 void
474 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m)
475 {
476 	struct ether_header *eh;
477 	struct mbuf *mtop = m;
478 	uint16_t ether_type;
479 	int hlen, af, hdrsize;
480 	void *hdr;
481 
482 	KASSERT((mtop->m_flags & M_PKTHDR) != 0);
483 
484 	hlen = ETHER_HDR_LEN;
485 	eh = mtod(m, struct ether_header *);
486 
487 	ether_type = htons(eh->ether_type);
488 
489 	if (ether_type < ETHERMTU) {
490 		/* LLC/SNAP */
491 		struct llc *llc = (struct llc *)(eh + 1);
492 		hlen += 8;
493 
494 		if (m->m_len < hlen ||
495 		    llc->llc_dsap != LLC_SNAP_LSAP ||
496 		    llc->llc_ssap != LLC_SNAP_LSAP ||
497 		    llc->llc_control != LLC_UI) {
498 			/* Not SNAP. */
499 			goto bad;
500 		}
501 
502 		ether_type = htons(llc->llc_un.type_snap.ether_type);
503 	}
504 
505 	switch (ether_type) {
506 	case ETHERTYPE_IP:
507 		af = AF_INET;
508 		hdrsize = 20;		/* sizeof(struct ip) */
509 		break;
510 
511 	case ETHERTYPE_IPV6:
512 		af = AF_INET6;
513 		hdrsize = 40;		/* sizeof(struct ip6_hdr) */
514 		break;
515 
516 	default:
517 		af = AF_UNSPEC;
518 		hdrsize = 0;
519 		break;
520 	}
521 
522 	while (m->m_len <= hlen) {
523 		hlen -= m->m_len;
524 		m = m->m_next;
525 		if (m == NULL)
526 			goto bad;
527 	}
528 
529 	if (m->m_len < (hlen + hdrsize)) {
530 		/*
531 		 * protocol header not in a single mbuf.
532 		 * We can't cope with this situation right
533 		 * now (but it shouldn't ever happen, really, anyhow).
534 		 */
535 #ifdef DEBUG
536 		printf("altq_etherclassify: headers span multiple mbufs: "
537 		    "%d < %d\n", m->m_len, (hlen + hdrsize));
538 #endif
539 		goto bad;
540 	}
541 
542 	m->m_data += hlen;
543 	m->m_len -= hlen;
544 
545 	hdr = mtod(m, void *);
546 
547 	if (ALTQ_NEEDS_CLASSIFY(ifq)) {
548 		mtop->m_pkthdr.pattr_class =
549 		    (*ifq->altq_classify)(ifq->altq_clfier, m, af);
550 	}
551 	mtop->m_pkthdr.pattr_af = af;
552 	mtop->m_pkthdr.pattr_hdr = hdr;
553 
554 	m->m_data -= hlen;
555 	m->m_len += hlen;
556 
557 	return;
558 
559 bad:
560 	mtop->m_pkthdr.pattr_class = NULL;
561 	mtop->m_pkthdr.pattr_hdr = NULL;
562 	mtop->m_pkthdr.pattr_af = AF_UNSPEC;
563 }
564 #endif /* ALTQ */
565 
566 /*
567  * Process a received Ethernet packet;
568  * the packet is in the mbuf chain m with
569  * the ether header.
570  */
571 void
572 ether_input(struct ifnet *ifp, struct mbuf *m)
573 {
574 	struct ethercom *ec = (struct ethercom *) ifp;
575 	pktqueue_t *pktq = NULL;
576 	struct ifqueue *inq = NULL;
577 	uint16_t etype;
578 	struct ether_header *eh;
579 	size_t ehlen;
580 	static int earlypkts;
581 	int isr = 0;
582 #if defined (LLC) || defined(NETATALK)
583 	struct llc *l;
584 #endif
585 
586 	KASSERT(!cpu_intr_p());
587 	KASSERT((m->m_flags & M_PKTHDR) != 0);
588 
589 	if ((ifp->if_flags & IFF_UP) == 0) {
590 		m_freem(m);
591 		return;
592 	}
593 	if (m->m_len < sizeof(*eh)) {
594 		m = m_pullup(m, sizeof(*eh));
595 		if (m == NULL)
596 			return;
597 	}
598 
599 #ifdef MBUFTRACE
600 	m_claimm(m, &ec->ec_rx_mowner);
601 #endif
602 	eh = mtod(m, struct ether_header *);
603 	etype = ntohs(eh->ether_type);
604 	ehlen = sizeof(*eh);
605 
606 	if (__predict_false(earlypkts < 100 ||
607 		entropy_epoch() == (unsigned)-1)) {
608 		rnd_add_data(NULL, eh, ehlen, 0);
609 		earlypkts++;
610 	}
611 
612 	/*
613 	 * Determine if the packet is within its size limits. For MPLS the
614 	 * header length is variable, so we skip the check.
615 	 */
616 	if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len >
617 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
618 #ifdef DIAGNOSTIC
619 		mutex_enter(&bigpktpps_lock);
620 		if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count,
621 		    bigpktppslim)) {
622 			printf("%s: discarding oversize frame (len=%d)\n",
623 			    ifp->if_xname, m->m_pkthdr.len);
624 		}
625 		mutex_exit(&bigpktpps_lock);
626 #endif
627 		if_statinc(ifp, if_iqdrops);
628 		m_freem(m);
629 		return;
630 	}
631 
632 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
633 		/*
634 		 * If this is not a simplex interface, drop the packet
635 		 * if it came from us.
636 		 */
637 		if ((ifp->if_flags & IFF_SIMPLEX) == 0 &&
638 		    memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost,
639 		    ETHER_ADDR_LEN) == 0) {
640 			m_freem(m);
641 			return;
642 		}
643 
644 		if (memcmp(etherbroadcastaddr,
645 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
646 			m->m_flags |= M_BCAST;
647 		else
648 			m->m_flags |= M_MCAST;
649 		if_statinc(ifp, if_imcasts);
650 	}
651 
652 	/* If the CRC is still on the packet, trim it off. */
653 	if (m->m_flags & M_HASFCS) {
654 		m_adj(m, -ETHER_CRC_LEN);
655 		m->m_flags &= ~M_HASFCS;
656 	}
657 
658 	if_statadd(ifp, if_ibytes, m->m_pkthdr.len);
659 
660 #if NCARP > 0
661 	if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) {
662 		/*
663 		 * Clear M_PROMISC, in case the packet comes from a
664 		 * vlan.
665 		 */
666 		m->m_flags &= ~M_PROMISC;
667 		if (carp_input(m, (uint8_t *)&eh->ether_shost,
668 		    (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0)
669 			return;
670 	}
671 #endif
672 
673 	if ((m->m_flags & (M_BCAST | M_MCAST | M_PROMISC)) == 0 &&
674 	    (ifp->if_flags & IFF_PROMISC) != 0 &&
675 	    memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost,
676 	     ETHER_ADDR_LEN) != 0) {
677 		m->m_flags |= M_PROMISC;
678 	}
679 
680 	if ((m->m_flags & M_PROMISC) == 0) {
681 		if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
682 			return;
683 		if (m == NULL)
684 			return;
685 
686 		eh = mtod(m, struct ether_header *);
687 		etype = ntohs(eh->ether_type);
688 	}
689 
690 #if NAGR > 0
691 	if (ifp->if_agrprivate &&
692 	    __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) {
693 		m->m_flags &= ~M_PROMISC;
694 		agr_input(ifp, m);
695 		return;
696 	}
697 #endif
698 
699 	/*
700 	 * If VLANs are configured on the interface, check to
701 	 * see if the device performed the decapsulation and
702 	 * provided us with the tag.
703 	 */
704 	if (ec->ec_nvlans && vlan_has_tag(m)) {
705 #if NVLAN > 0
706 		/*
707 		 * vlan_input() will either recursively call ether_input()
708 		 * or drop the packet.
709 		 */
710 		vlan_input(ifp, m);
711 #else
712 		m_freem(m);
713 #endif
714 		return;
715 	}
716 
717 	/*
718 	 * Handle protocols that expect to have the Ethernet header
719 	 * (and possibly FCS) intact.
720 	 */
721 	switch (etype) {
722 	case ETHERTYPE_VLAN: {
723 		struct ether_vlan_header *evl = (void *)eh;
724 
725 		/*
726 		 * If there is a tag of 0, then the VLAN header was probably
727 		 * just being used to store the priority.  Extract the ether
728 		 * type, and if IP or IPV6, let them deal with it.
729 		 */
730 		if (m->m_len >= sizeof(*evl) &&
731 		    EVL_VLANOFTAG(evl->evl_tag) == 0) {
732 			etype = ntohs(evl->evl_proto);
733 			ehlen = sizeof(*evl);
734 			if ((m->m_flags & M_PROMISC) == 0 &&
735 			    (etype == ETHERTYPE_IP ||
736 			     etype == ETHERTYPE_IPV6))
737 				break;
738 		}
739 
740 #if NVLAN > 0
741 		/*
742 		 * vlan_input() will either recursively call ether_input()
743 		 * or drop the packet.
744 		 */
745 		if (ec->ec_nvlans != 0)
746 			vlan_input(ifp, m);
747 		else
748 #endif
749 			m_freem(m);
750 
751 		return;
752 	}
753 
754 #if NPPPOE > 0
755 	case ETHERTYPE_PPPOEDISC:
756 		pppoedisc_input(ifp, m);
757 		return;
758 
759 	case ETHERTYPE_PPPOE:
760 		pppoe_input(ifp, m);
761 		return;
762 #endif
763 
764 	case ETHERTYPE_SLOWPROTOCOLS: {
765 		uint8_t subtype;
766 
767 		if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype)) {
768 			m_freem(m);
769 			return;
770 		}
771 
772 		m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype);
773 		switch (subtype) {
774 #if NAGR > 0
775 		case SLOWPROTOCOLS_SUBTYPE_LACP:
776 			if (ifp->if_agrprivate) {
777 				ieee8023ad_lacp_input(ifp, m);
778 				return;
779 			}
780 			break;
781 
782 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
783 			if (ifp->if_agrprivate) {
784 				ieee8023ad_marker_input(ifp, m);
785 				return;
786 			}
787 			break;
788 #endif
789 
790 		default:
791 			if (subtype == 0 || subtype > 10) {
792 				/* illegal value */
793 				m_freem(m);
794 				return;
795 			}
796 			/* unknown subtype */
797 			break;
798 		}
799 	}
800 	/* FALLTHROUGH */
801 	default:
802 		if (m->m_flags & M_PROMISC) {
803 			m_freem(m);
804 			return;
805 		}
806 	}
807 
808 	/* If the CRC is still on the packet, trim it off. */
809 	if (m->m_flags & M_HASFCS) {
810 		m_adj(m, -ETHER_CRC_LEN);
811 		m->m_flags &= ~M_HASFCS;
812 	}
813 
814 	if (etype > ETHERMTU + sizeof(struct ether_header)) {
815 		/* Strip off the Ethernet header. */
816 		m_adj(m, ehlen);
817 
818 		switch (etype) {
819 #ifdef INET
820 		case ETHERTYPE_IP:
821 #ifdef GATEWAY
822 			if (ipflow_fastforward(m))
823 				return;
824 #endif
825 			pktq = ip_pktq;
826 			break;
827 
828 		case ETHERTYPE_ARP:
829 			isr = NETISR_ARP;
830 			inq = &arpintrq;
831 			break;
832 
833 		case ETHERTYPE_REVARP:
834 			revarpinput(m);	/* XXX queue? */
835 			return;
836 #endif
837 
838 #ifdef INET6
839 		case ETHERTYPE_IPV6:
840 			if (__predict_false(!in6_present)) {
841 				m_freem(m);
842 				return;
843 			}
844 #ifdef GATEWAY
845 			if (ip6flow_fastforward(&m))
846 				return;
847 #endif
848 			pktq = ip6_pktq;
849 			break;
850 #endif
851 
852 #ifdef NETATALK
853 		case ETHERTYPE_ATALK:
854 			isr = NETISR_ATALK;
855 			inq = &atintrq1;
856 			break;
857 
858 		case ETHERTYPE_AARP:
859 			aarpinput(ifp, m); /* XXX queue? */
860 			return;
861 #endif
862 
863 #ifdef MPLS
864 		case ETHERTYPE_MPLS:
865 			isr = NETISR_MPLS;
866 			inq = &mplsintrq;
867 			break;
868 #endif
869 
870 		default:
871 			m_freem(m);
872 			return;
873 		}
874 	} else {
875 		KASSERT(ehlen == sizeof(*eh));
876 #if defined (LLC) || defined (NETATALK)
877 		if (m->m_len < sizeof(*eh) + sizeof(struct llc)) {
878 			goto dropanyway;
879 		}
880 		l = (struct llc *)(eh+1);
881 
882 		switch (l->llc_dsap) {
883 #ifdef NETATALK
884 		case LLC_SNAP_LSAP:
885 			switch (l->llc_control) {
886 			case LLC_UI:
887 				if (l->llc_ssap != LLC_SNAP_LSAP) {
888 					goto dropanyway;
889 				}
890 
891 				if (memcmp(&(l->llc_snap_org_code)[0],
892 				    at_org_code, sizeof(at_org_code)) == 0 &&
893 				    ntohs(l->llc_snap_ether_type) ==
894 				    ETHERTYPE_ATALK) {
895 					inq = &atintrq2;
896 					m_adj(m, sizeof(struct ether_header)
897 					    + sizeof(struct llc));
898 					isr = NETISR_ATALK;
899 					break;
900 				}
901 
902 				if (memcmp(&(l->llc_snap_org_code)[0],
903 				    aarp_org_code,
904 				    sizeof(aarp_org_code)) == 0 &&
905 				    ntohs(l->llc_snap_ether_type) ==
906 				    ETHERTYPE_AARP) {
907 					m_adj(m, sizeof(struct ether_header)
908 					    + sizeof(struct llc));
909 					aarpinput(ifp, m); /* XXX queue? */
910 					return;
911 				}
912 
913 			default:
914 				goto dropanyway;
915 			}
916 			break;
917 #endif
918 		dropanyway:
919 		default:
920 			m_freem(m);
921 			return;
922 		}
923 #else /* LLC || NETATALK */
924 		m_freem(m);
925 		return;
926 #endif /* LLC || NETATALK */
927 	}
928 
929 	if (__predict_true(pktq)) {
930 #ifdef NET_MPSAFE
931 		const u_int h = curcpu()->ci_index;
932 #else
933 		const uint32_t h = pktq_rps_hash(m);
934 #endif
935 		if (__predict_false(!pktq_enqueue(pktq, m, h))) {
936 			m_freem(m);
937 		}
938 		return;
939 	}
940 
941 	if (__predict_false(!inq)) {
942 		/* Should not happen. */
943 		m_freem(m);
944 		return;
945 	}
946 
947 	IFQ_LOCK(inq);
948 	if (IF_QFULL(inq)) {
949 		IF_DROP(inq);
950 		IFQ_UNLOCK(inq);
951 		m_freem(m);
952 	} else {
953 		IF_ENQUEUE(inq, m);
954 		IFQ_UNLOCK(inq);
955 		schednetisr(isr);
956 	}
957 }
958 
959 /*
960  * Convert Ethernet address to printable (loggable) representation.
961  */
962 char *
963 ether_sprintf(const u_char *ap)
964 {
965 	static char etherbuf[3 * ETHER_ADDR_LEN];
966 	return ether_snprintf(etherbuf, sizeof(etherbuf), ap);
967 }
968 
969 char *
970 ether_snprintf(char *buf, size_t len, const u_char *ap)
971 {
972 	char *cp = buf;
973 	size_t i;
974 
975 	for (i = 0; i < len / 3; i++) {
976 		*cp++ = hexdigits[*ap >> 4];
977 		*cp++ = hexdigits[*ap++ & 0xf];
978 		*cp++ = ':';
979 	}
980 	*--cp = '\0';
981 	return buf;
982 }
983 
984 /*
985  * Perform common duties while attaching to interface list
986  */
987 void
988 ether_ifattach(struct ifnet *ifp, const uint8_t *lla)
989 {
990 	struct ethercom *ec = (struct ethercom *)ifp;
991 
992 	ifp->if_type = IFT_ETHER;
993 	ifp->if_hdrlen = ETHER_HDR_LEN;
994 	ifp->if_dlt = DLT_EN10MB;
995 	ifp->if_mtu = ETHERMTU;
996 	ifp->if_output = ether_output;
997 	ifp->_if_input = ether_input;
998 	if (ifp->if_baudrate == 0)
999 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
1000 
1001 	if (lla != NULL)
1002 		if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla));
1003 
1004 	LIST_INIT(&ec->ec_multiaddrs);
1005 	SIMPLEQ_INIT(&ec->ec_vids);
1006 	ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
1007 	ec->ec_flags = 0;
1008 	ifp->if_broadcastaddr = etherbroadcastaddr;
1009 	bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header));
1010 #ifdef MBUFTRACE
1011 	mowner_init_owner(&ec->ec_tx_mowner, ifp->if_xname, "tx");
1012 	mowner_init_owner(&ec->ec_rx_mowner, ifp->if_xname, "rx");
1013 	MOWNER_ATTACH(&ec->ec_tx_mowner);
1014 	MOWNER_ATTACH(&ec->ec_rx_mowner);
1015 	ifp->if_mowner = &ec->ec_tx_mowner;
1016 #endif
1017 }
1018 
1019 void
1020 ether_ifdetach(struct ifnet *ifp)
1021 {
1022 	struct ethercom *ec = (void *) ifp;
1023 	struct ether_multi *enm;
1024 
1025 	IFNET_ASSERT_UNLOCKED(ifp);
1026 	/*
1027 	 * Prevent further calls to ioctl (for example turning off
1028 	 * promiscuous mode from the bridge code), which eventually can
1029 	 * call if_init() which can cause panics because the interface
1030 	 * is in the process of being detached. Return device not configured
1031 	 * instead.
1032 	 */
1033 	ifp->if_ioctl = __FPTRCAST(int (*)(struct ifnet *, u_long, void *),
1034 	    enxio);
1035 
1036 #if NBRIDGE > 0
1037 	if (ifp->if_bridge)
1038 		bridge_ifdetach(ifp);
1039 #endif
1040 	bpf_detach(ifp);
1041 #if NVLAN > 0
1042 	if (ec->ec_nvlans)
1043 		vlan_ifdetach(ifp);
1044 #endif
1045 
1046 	ETHER_LOCK(ec);
1047 	KASSERT(ec->ec_nvlans == 0);
1048 	while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) {
1049 		LIST_REMOVE(enm, enm_list);
1050 		kmem_free(enm, sizeof(*enm));
1051 		ec->ec_multicnt--;
1052 	}
1053 	ETHER_UNLOCK(ec);
1054 
1055 	mutex_obj_free(ec->ec_lock);
1056 	ec->ec_lock = NULL;
1057 
1058 	ifp->if_mowner = NULL;
1059 	MOWNER_DETACH(&ec->ec_rx_mowner);
1060 	MOWNER_DETACH(&ec->ec_tx_mowner);
1061 }
1062 
1063 #if 0
1064 /*
1065  * This is for reference.  We have a table-driven version
1066  * of the little-endian crc32 generator, which is faster
1067  * than the double-loop.
1068  */
1069 uint32_t
1070 ether_crc32_le(const uint8_t *buf, size_t len)
1071 {
1072 	uint32_t c, crc, carry;
1073 	size_t i, j;
1074 
1075 	crc = 0xffffffffU;	/* initial value */
1076 
1077 	for (i = 0; i < len; i++) {
1078 		c = buf[i];
1079 		for (j = 0; j < 8; j++) {
1080 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1081 			crc >>= 1;
1082 			c >>= 1;
1083 			if (carry)
1084 				crc = (crc ^ ETHER_CRC_POLY_LE);
1085 		}
1086 	}
1087 
1088 	return (crc);
1089 }
1090 #else
1091 uint32_t
1092 ether_crc32_le(const uint8_t *buf, size_t len)
1093 {
1094 	static const uint32_t crctab[] = {
1095 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1096 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1097 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1098 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1099 	};
1100 	uint32_t crc;
1101 	size_t i;
1102 
1103 	crc = 0xffffffffU;	/* initial value */
1104 
1105 	for (i = 0; i < len; i++) {
1106 		crc ^= buf[i];
1107 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1108 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1109 	}
1110 
1111 	return (crc);
1112 }
1113 #endif
1114 
1115 uint32_t
1116 ether_crc32_be(const uint8_t *buf, size_t len)
1117 {
1118 	uint32_t c, crc, carry;
1119 	size_t i, j;
1120 
1121 	crc = 0xffffffffU;	/* initial value */
1122 
1123 	for (i = 0; i < len; i++) {
1124 		c = buf[i];
1125 		for (j = 0; j < 8; j++) {
1126 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1127 			crc <<= 1;
1128 			c >>= 1;
1129 			if (carry)
1130 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1131 		}
1132 	}
1133 
1134 	return (crc);
1135 }
1136 
1137 #ifdef INET
1138 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] =
1139     { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 };
1140 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] =
1141     { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff };
1142 #endif
1143 #ifdef INET6
1144 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] =
1145     { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 };
1146 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] =
1147     { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff };
1148 #endif
1149 
1150 /*
1151  * ether_aton implementation, not using a static buffer.
1152  */
1153 int
1154 ether_aton_r(u_char *dest, size_t len, const char *str)
1155 {
1156 	const u_char *cp = (const void *)str;
1157 	u_char *ep;
1158 
1159 #define atox(c)	(((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10))
1160 
1161 	if (len < ETHER_ADDR_LEN)
1162 		return ENOSPC;
1163 
1164 	ep = dest + ETHER_ADDR_LEN;
1165 
1166 	while (*cp) {
1167 		if (!isxdigit(*cp))
1168 			return EINVAL;
1169 
1170 		*dest = atox(*cp);
1171 		cp++;
1172 		if (isxdigit(*cp)) {
1173 			*dest = (*dest << 4) | atox(*cp);
1174 			cp++;
1175 		}
1176 		dest++;
1177 
1178 		if (dest == ep)
1179 			return (*cp == '\0') ? 0 : ENAMETOOLONG;
1180 
1181 		switch (*cp) {
1182 		case ':':
1183 		case '-':
1184 		case '.':
1185 			cp++;
1186 			break;
1187 		}
1188 	}
1189 	return ENOBUFS;
1190 }
1191 
1192 /*
1193  * Convert a sockaddr into an Ethernet address or range of Ethernet
1194  * addresses.
1195  */
1196 int
1197 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN],
1198     uint8_t addrhi[ETHER_ADDR_LEN])
1199 {
1200 #ifdef INET
1201 	const struct sockaddr_in *sin;
1202 #endif
1203 #ifdef INET6
1204 	const struct sockaddr_in6 *sin6;
1205 #endif
1206 
1207 	switch (sa->sa_family) {
1208 
1209 	case AF_UNSPEC:
1210 		memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN);
1211 		memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1212 		break;
1213 
1214 #ifdef INET
1215 	case AF_INET:
1216 		sin = satocsin(sa);
1217 		if (sin->sin_addr.s_addr == INADDR_ANY) {
1218 			/*
1219 			 * An IP address of INADDR_ANY means listen to
1220 			 * or stop listening to all of the Ethernet
1221 			 * multicast addresses used for IP.
1222 			 * (This is for the sake of IP multicast routers.)
1223 			 */
1224 			memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN);
1225 			memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN);
1226 		} else {
1227 			ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo);
1228 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1229 		}
1230 		break;
1231 #endif
1232 #ifdef INET6
1233 	case AF_INET6:
1234 		sin6 = satocsin6(sa);
1235 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1236 			/*
1237 			 * An IP6 address of 0 means listen to or stop
1238 			 * listening to all of the Ethernet multicast
1239 			 * address used for IP6.
1240 			 * (This is used for multicast routers.)
1241 			 */
1242 			memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN);
1243 			memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN);
1244 		} else {
1245 			ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo);
1246 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1247 		}
1248 		break;
1249 #endif
1250 
1251 	default:
1252 		return EAFNOSUPPORT;
1253 	}
1254 	return 0;
1255 }
1256 
1257 /*
1258  * Add an Ethernet multicast address or range of addresses to the list for a
1259  * given interface.
1260  */
1261 int
1262 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec)
1263 {
1264 	struct ether_multi *enm, *_enm;
1265 	u_char addrlo[ETHER_ADDR_LEN];
1266 	u_char addrhi[ETHER_ADDR_LEN];
1267 	int error = 0;
1268 
1269 	/* Allocate out of lock */
1270 	enm = kmem_alloc(sizeof(*enm), KM_SLEEP);
1271 
1272 	ETHER_LOCK(ec);
1273 	error = ether_multiaddr(sa, addrlo, addrhi);
1274 	if (error != 0)
1275 		goto out;
1276 
1277 	/*
1278 	 * Verify that we have valid Ethernet multicast addresses.
1279 	 */
1280 	if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) {
1281 		error = EINVAL;
1282 		goto out;
1283 	}
1284 
1285 	/*
1286 	 * See if the address range is already in the list.
1287 	 */
1288 	_enm = ether_lookup_multi(addrlo, addrhi, ec);
1289 	if (_enm != NULL) {
1290 		/*
1291 		 * Found it; just increment the reference count.
1292 		 */
1293 		++_enm->enm_refcount;
1294 		error = 0;
1295 		goto out;
1296 	}
1297 
1298 	/*
1299 	 * Link a new multicast record into the interface's multicast list.
1300 	 */
1301 	memcpy(enm->enm_addrlo, addrlo, ETHER_ADDR_LEN);
1302 	memcpy(enm->enm_addrhi, addrhi, ETHER_ADDR_LEN);
1303 	enm->enm_refcount = 1;
1304 	LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list);
1305 	ec->ec_multicnt++;
1306 
1307 	/*
1308 	 * Return ENETRESET to inform the driver that the list has changed
1309 	 * and its reception filter should be adjusted accordingly.
1310 	 */
1311 	error = ENETRESET;
1312 	enm = NULL;
1313 
1314 out:
1315 	ETHER_UNLOCK(ec);
1316 	if (enm != NULL)
1317 		kmem_free(enm, sizeof(*enm));
1318 	return error;
1319 }
1320 
1321 /*
1322  * Delete a multicast address record.
1323  */
1324 int
1325 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec)
1326 {
1327 	struct ether_multi *enm;
1328 	u_char addrlo[ETHER_ADDR_LEN];
1329 	u_char addrhi[ETHER_ADDR_LEN];
1330 	int error;
1331 
1332 	ETHER_LOCK(ec);
1333 	error = ether_multiaddr(sa, addrlo, addrhi);
1334 	if (error != 0)
1335 		goto error;
1336 
1337 	/*
1338 	 * Look up the address in our list.
1339 	 */
1340 	enm = ether_lookup_multi(addrlo, addrhi, ec);
1341 	if (enm == NULL) {
1342 		error = ENXIO;
1343 		goto error;
1344 	}
1345 	if (--enm->enm_refcount != 0) {
1346 		/*
1347 		 * Still some claims to this record.
1348 		 */
1349 		error = 0;
1350 		goto error;
1351 	}
1352 
1353 	/*
1354 	 * No remaining claims to this record; unlink and free it.
1355 	 */
1356 	LIST_REMOVE(enm, enm_list);
1357 	ec->ec_multicnt--;
1358 	ETHER_UNLOCK(ec);
1359 	kmem_free(enm, sizeof(*enm));
1360 
1361 	/*
1362 	 * Return ENETRESET to inform the driver that the list has changed
1363 	 * and its reception filter should be adjusted accordingly.
1364 	 */
1365 	return ENETRESET;
1366 
1367 error:
1368 	ETHER_UNLOCK(ec);
1369 	return error;
1370 }
1371 
1372 void
1373 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb)
1374 {
1375 	ec->ec_ifflags_cb = cb;
1376 }
1377 
1378 void
1379 ether_set_vlan_cb(struct ethercom *ec, ether_vlancb_t cb)
1380 {
1381 
1382 	ec->ec_vlan_cb = cb;
1383 }
1384 
1385 static int
1386 ether_ioctl_reinit(struct ethercom *ec)
1387 {
1388 	struct ifnet *ifp = &ec->ec_if;
1389 	int error;
1390 
1391 	switch (ifp->if_flags & (IFF_UP | IFF_RUNNING)) {
1392 	case IFF_RUNNING:
1393 		/*
1394 		 * If interface is marked down and it is running,
1395 		 * then stop and disable it.
1396 		 */
1397 		(*ifp->if_stop)(ifp, 1);
1398 		break;
1399 	case IFF_UP:
1400 		/*
1401 		 * If interface is marked up and it is stopped, then
1402 		 * start it.
1403 		 */
1404 		return (*ifp->if_init)(ifp);
1405 	case IFF_UP | IFF_RUNNING:
1406 		error = 0;
1407 		if (ec->ec_ifflags_cb != NULL) {
1408 			error = (*ec->ec_ifflags_cb)(ec);
1409 			if (error == ENETRESET) {
1410 				/*
1411 				 * Reset the interface to pick up
1412 				 * changes in any other flags that
1413 				 * affect the hardware state.
1414 				 */
1415 				return (*ifp->if_init)(ifp);
1416 			}
1417 		} else
1418 			error = (*ifp->if_init)(ifp);
1419 		return error;
1420 	case 0:
1421 		break;
1422 	}
1423 
1424 	return 0;
1425 }
1426 
1427 /*
1428  * Common ioctls for Ethernet interfaces.  Note, we must be
1429  * called at splnet().
1430  */
1431 int
1432 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data)
1433 {
1434 	struct ethercom *ec = (void *)ifp;
1435 	struct eccapreq *eccr;
1436 	struct ifreq *ifr = (struct ifreq *)data;
1437 	struct if_laddrreq *iflr = data;
1438 	const struct sockaddr_dl *sdl;
1439 	static const uint8_t zero[ETHER_ADDR_LEN];
1440 	int error;
1441 
1442 	switch (cmd) {
1443 	case SIOCINITIFADDR:
1444 	    {
1445 		struct ifaddr *ifa = (struct ifaddr *)data;
1446 		if (ifa->ifa_addr->sa_family != AF_LINK
1447 		    && (ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
1448 		       (IFF_UP | IFF_RUNNING)) {
1449 			ifp->if_flags |= IFF_UP;
1450 			if ((error = (*ifp->if_init)(ifp)) != 0)
1451 				return error;
1452 		}
1453 #ifdef INET
1454 		if (ifa->ifa_addr->sa_family == AF_INET)
1455 			arp_ifinit(ifp, ifa);
1456 #endif
1457 		return 0;
1458 	    }
1459 
1460 	case SIOCSIFMTU:
1461 	    {
1462 		int maxmtu;
1463 
1464 		if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU)
1465 			maxmtu = ETHERMTU_JUMBO;
1466 		else
1467 			maxmtu = ETHERMTU;
1468 
1469 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu)
1470 			return EINVAL;
1471 		else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET)
1472 			return error;
1473 		else if (ifp->if_flags & IFF_UP) {
1474 			/* Make sure the device notices the MTU change. */
1475 			return (*ifp->if_init)(ifp);
1476 		} else
1477 			return 0;
1478 	    }
1479 
1480 	case SIOCSIFFLAGS:
1481 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
1482 			return error;
1483 		return ether_ioctl_reinit(ec);
1484 	case SIOCGIFFLAGS:
1485 		error = ifioctl_common(ifp, cmd, data);
1486 		if (error == 0) {
1487 			/* Set IFF_ALLMULTI for backcompat */
1488 			ifr->ifr_flags |= (ec->ec_flags & ETHER_F_ALLMULTI) ?
1489 			    IFF_ALLMULTI : 0;
1490 		}
1491 		return error;
1492 	case SIOCGETHERCAP:
1493 		eccr = (struct eccapreq *)data;
1494 		eccr->eccr_capabilities = ec->ec_capabilities;
1495 		eccr->eccr_capenable = ec->ec_capenable;
1496 		return 0;
1497 	case SIOCSETHERCAP:
1498 		eccr = (struct eccapreq *)data;
1499 		if ((eccr->eccr_capenable & ~ec->ec_capabilities) != 0)
1500 			return EINVAL;
1501 		if (eccr->eccr_capenable == ec->ec_capenable)
1502 			return 0;
1503 #if 0 /* notyet */
1504 		ec->ec_capenable = (ec->ec_capenable & ETHERCAP_CANTCHANGE)
1505 		    | (eccr->eccr_capenable & ~ETHERCAP_CANTCHANGE);
1506 #else
1507 		ec->ec_capenable = eccr->eccr_capenable;
1508 #endif
1509 		return ether_ioctl_reinit(ec);
1510 	case SIOCADDMULTI:
1511 		return ether_addmulti(ifreq_getaddr(cmd, ifr), ec);
1512 	case SIOCDELMULTI:
1513 		return ether_delmulti(ifreq_getaddr(cmd, ifr), ec);
1514 	case SIOCSIFMEDIA:
1515 	case SIOCGIFMEDIA:
1516 		if (ec->ec_mii != NULL)
1517 			return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media,
1518 			    cmd);
1519 		else if (ec->ec_ifmedia != NULL)
1520 			return ifmedia_ioctl(ifp, ifr, ec->ec_ifmedia, cmd);
1521 		else
1522 			return ENOTTY;
1523 		break;
1524 	case SIOCALIFADDR:
1525 		sdl = satocsdl(sstocsa(&iflr->addr));
1526 		if (sdl->sdl_family != AF_LINK)
1527 			;
1528 		else if (ETHER_IS_MULTICAST(CLLADDR(sdl)))
1529 			return EINVAL;
1530 		else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0)
1531 			return EINVAL;
1532 		/*FALLTHROUGH*/
1533 	default:
1534 		return ifioctl_common(ifp, cmd, data);
1535 	}
1536 	return 0;
1537 }
1538 
1539 /*
1540  * Enable/disable passing VLAN packets if the parent interface supports it.
1541  * Return:
1542  * 	 0: Ok
1543  *	-1: Parent interface does not support vlans
1544  *	>0: Error
1545  */
1546 int
1547 ether_enable_vlan_mtu(struct ifnet *ifp)
1548 {
1549 	int error;
1550 	struct ethercom *ec = (void *)ifp;
1551 
1552 	/* Parent does not support VLAN's */
1553 	if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0)
1554 		return -1;
1555 
1556 	/*
1557 	 * Parent supports the VLAN_MTU capability,
1558 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames;
1559 	 * enable it.
1560 	 */
1561 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1562 
1563 	/* Interface is down, defer for later */
1564 	if ((ifp->if_flags & IFF_UP) == 0)
1565 		return 0;
1566 
1567 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1568 		return 0;
1569 
1570 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1571 	return error;
1572 }
1573 
1574 int
1575 ether_disable_vlan_mtu(struct ifnet *ifp)
1576 {
1577 	int error;
1578 	struct ethercom *ec = (void *)ifp;
1579 
1580 	/* We still have VLAN's, defer for later */
1581 	if (ec->ec_nvlans != 0)
1582 		return 0;
1583 
1584 	/* Parent does not support VLAB's, nothing to do. */
1585 	if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0)
1586 		return -1;
1587 
1588 	/*
1589 	 * Disable Tx/Rx of VLAN-sized frames.
1590 	 */
1591 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1592 
1593 	/* Interface is down, defer for later */
1594 	if ((ifp->if_flags & IFF_UP) == 0)
1595 		return 0;
1596 
1597 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1598 		return 0;
1599 
1600 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1601 	return error;
1602 }
1603 
1604 static int
1605 ether_multicast_sysctl(SYSCTLFN_ARGS)
1606 {
1607 	struct ether_multi *enm;
1608 	struct ifnet *ifp;
1609 	struct ethercom *ec;
1610 	int error = 0;
1611 	size_t written;
1612 	struct psref psref;
1613 	int bound;
1614 	unsigned int multicnt;
1615 	struct ether_multi_sysctl *addrs;
1616 	int i;
1617 
1618 	if (namelen != 1)
1619 		return EINVAL;
1620 
1621 	bound = curlwp_bind();
1622 	ifp = if_get_byindex(name[0], &psref);
1623 	if (ifp == NULL) {
1624 		error = ENODEV;
1625 		goto out;
1626 	}
1627 	if (ifp->if_type != IFT_ETHER) {
1628 		if_put(ifp, &psref);
1629 		*oldlenp = 0;
1630 		goto out;
1631 	}
1632 	ec = (struct ethercom *)ifp;
1633 
1634 	if (oldp == NULL) {
1635 		if_put(ifp, &psref);
1636 		*oldlenp = ec->ec_multicnt * sizeof(*addrs);
1637 		goto out;
1638 	}
1639 
1640 	/*
1641 	 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which
1642 	 * is sleepable, while holding it. Copy data to a local buffer first
1643 	 * with the lock taken and then call sysctl_copyout without holding it.
1644 	 */
1645 retry:
1646 	multicnt = ec->ec_multicnt;
1647 
1648 	if (multicnt == 0) {
1649 		if_put(ifp, &psref);
1650 		*oldlenp = 0;
1651 		goto out;
1652 	}
1653 
1654 	addrs = kmem_zalloc(sizeof(*addrs) * multicnt, KM_SLEEP);
1655 
1656 	ETHER_LOCK(ec);
1657 	if (multicnt != ec->ec_multicnt) {
1658 		/* The number of multicast addresses has changed */
1659 		ETHER_UNLOCK(ec);
1660 		kmem_free(addrs, sizeof(*addrs) * multicnt);
1661 		goto retry;
1662 	}
1663 
1664 	i = 0;
1665 	LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) {
1666 		struct ether_multi_sysctl *addr = &addrs[i];
1667 		addr->enm_refcount = enm->enm_refcount;
1668 		memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN);
1669 		memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN);
1670 		i++;
1671 	}
1672 	ETHER_UNLOCK(ec);
1673 
1674 	error = 0;
1675 	written = 0;
1676 	for (i = 0; i < multicnt; i++) {
1677 		struct ether_multi_sysctl *addr = &addrs[i];
1678 
1679 		if (written + sizeof(*addr) > *oldlenp)
1680 			break;
1681 		error = sysctl_copyout(l, addr, oldp, sizeof(*addr));
1682 		if (error)
1683 			break;
1684 		written += sizeof(*addr);
1685 		oldp = (char *)oldp + sizeof(*addr);
1686 	}
1687 	kmem_free(addrs, sizeof(*addrs) * multicnt);
1688 
1689 	if_put(ifp, &psref);
1690 
1691 	*oldlenp = written;
1692 out:
1693 	curlwp_bindx(bound);
1694 	return error;
1695 }
1696 
1697 static void
1698 ether_sysctl_setup(struct sysctllog **clog)
1699 {
1700 	const struct sysctlnode *rnode = NULL;
1701 
1702 	sysctl_createv(clog, 0, NULL, &rnode,
1703 		       CTLFLAG_PERMANENT,
1704 		       CTLTYPE_NODE, "ether",
1705 		       SYSCTL_DESCR("Ethernet-specific information"),
1706 		       NULL, 0, NULL, 0,
1707 		       CTL_NET, CTL_CREATE, CTL_EOL);
1708 
1709 	sysctl_createv(clog, 0, &rnode, NULL,
1710 		       CTLFLAG_PERMANENT,
1711 		       CTLTYPE_NODE, "multicast",
1712 		       SYSCTL_DESCR("multicast addresses"),
1713 		       ether_multicast_sysctl, 0, NULL, 0,
1714 		       CTL_CREATE, CTL_EOL);
1715 }
1716 
1717 void
1718 etherinit(void)
1719 {
1720 
1721 #ifdef DIAGNOSTIC
1722 	mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET);
1723 #endif
1724 	ether_sysctl_setup(NULL);
1725 }
1726