xref: /netbsd-src/sys/net/if_ethersubr.c (revision ed75d7a867996c84cfa88e3b8906816277e957f7)
1 /*	$NetBSD: if_ethersubr.c,v 1.282 2020/01/29 04:11:35 thorpej 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 1.282 2020/01/29 04:11:35 thorpej 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 #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 || !rnd_initial_entropy)) {
607 		rnd_add_data(NULL, eh, ehlen, 0);
608 		earlypkts++;
609 	}
610 
611 	/*
612 	 * Determine if the packet is within its size limits. For MPLS the
613 	 * header length is variable, so we skip the check.
614 	 */
615 	if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len >
616 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
617 #ifdef DIAGNOSTIC
618 		mutex_enter(&bigpktpps_lock);
619 		if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count,
620 		    bigpktppslim)) {
621 			printf("%s: discarding oversize frame (len=%d)\n",
622 			    ifp->if_xname, m->m_pkthdr.len);
623 		}
624 		mutex_exit(&bigpktpps_lock);
625 #endif
626 		if_statinc(ifp, if_iqdrops);
627 		m_freem(m);
628 		return;
629 	}
630 
631 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
632 		/*
633 		 * If this is not a simplex interface, drop the packet
634 		 * if it came from us.
635 		 */
636 		if ((ifp->if_flags & IFF_SIMPLEX) == 0 &&
637 		    memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost,
638 		    ETHER_ADDR_LEN) == 0) {
639 			m_freem(m);
640 			return;
641 		}
642 
643 		if (memcmp(etherbroadcastaddr,
644 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
645 			m->m_flags |= M_BCAST;
646 		else
647 			m->m_flags |= M_MCAST;
648 		if_statinc(ifp, if_imcasts);
649 	}
650 
651 	/* If the CRC is still on the packet, trim it off. */
652 	if (m->m_flags & M_HASFCS) {
653 		m_adj(m, -ETHER_CRC_LEN);
654 		m->m_flags &= ~M_HASFCS;
655 	}
656 
657 	if_statadd(ifp, if_ibytes, m->m_pkthdr.len);
658 
659 #if NCARP > 0
660 	if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) {
661 		/*
662 		 * Clear M_PROMISC, in case the packet comes from a
663 		 * vlan.
664 		 */
665 		m->m_flags &= ~M_PROMISC;
666 		if (carp_input(m, (uint8_t *)&eh->ether_shost,
667 		    (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0)
668 			return;
669 	}
670 #endif
671 
672 	if ((m->m_flags & (M_BCAST | M_MCAST | M_PROMISC)) == 0 &&
673 	    (ifp->if_flags & IFF_PROMISC) != 0 &&
674 	    memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost,
675 	     ETHER_ADDR_LEN) != 0) {
676 		m->m_flags |= M_PROMISC;
677 	}
678 
679 	if ((m->m_flags & M_PROMISC) == 0) {
680 		if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
681 			return;
682 		if (m == NULL)
683 			return;
684 
685 		eh = mtod(m, struct ether_header *);
686 		etype = ntohs(eh->ether_type);
687 	}
688 
689 #if NAGR > 0
690 	if (ifp->if_agrprivate &&
691 	    __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) {
692 		m->m_flags &= ~M_PROMISC;
693 		agr_input(ifp, m);
694 		return;
695 	}
696 #endif
697 
698 	/*
699 	 * If VLANs are configured on the interface, check to
700 	 * see if the device performed the decapsulation and
701 	 * provided us with the tag.
702 	 */
703 	if (ec->ec_nvlans && vlan_has_tag(m)) {
704 #if NVLAN > 0
705 		/*
706 		 * vlan_input() will either recursively call ether_input()
707 		 * or drop the packet.
708 		 */
709 		vlan_input(ifp, m);
710 #else
711 		m_freem(m);
712 #endif
713 		return;
714 	}
715 
716 	/*
717 	 * Handle protocols that expect to have the Ethernet header
718 	 * (and possibly FCS) intact.
719 	 */
720 	switch (etype) {
721 	case ETHERTYPE_VLAN: {
722 		struct ether_vlan_header *evl = (void *)eh;
723 
724 		/*
725 		 * If there is a tag of 0, then the VLAN header was probably
726 		 * just being used to store the priority.  Extract the ether
727 		 * type, and if IP or IPV6, let them deal with it.
728 		 */
729 		if (m->m_len >= sizeof(*evl) &&
730 		    EVL_VLANOFTAG(evl->evl_tag) == 0) {
731 			etype = ntohs(evl->evl_proto);
732 			ehlen = sizeof(*evl);
733 			if ((m->m_flags & M_PROMISC) == 0 &&
734 			    (etype == ETHERTYPE_IP ||
735 			     etype == ETHERTYPE_IPV6))
736 				break;
737 		}
738 
739 #if NVLAN > 0
740 		/*
741 		 * vlan_input() will either recursively call ether_input()
742 		 * or drop the packet.
743 		 */
744 		if (ec->ec_nvlans != 0)
745 			vlan_input(ifp, m);
746 		else
747 #endif
748 			m_freem(m);
749 
750 		return;
751 	}
752 
753 #if NPPPOE > 0
754 	case ETHERTYPE_PPPOEDISC:
755 		pppoedisc_input(ifp, m);
756 		return;
757 
758 	case ETHERTYPE_PPPOE:
759 		pppoe_input(ifp, m);
760 		return;
761 #endif
762 
763 	case ETHERTYPE_SLOWPROTOCOLS: {
764 		uint8_t subtype;
765 
766 		if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype)) {
767 			m_freem(m);
768 			return;
769 		}
770 
771 		m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype);
772 		switch (subtype) {
773 #if NAGR > 0
774 		case SLOWPROTOCOLS_SUBTYPE_LACP:
775 			if (ifp->if_agrprivate) {
776 				ieee8023ad_lacp_input(ifp, m);
777 				return;
778 			}
779 			break;
780 
781 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
782 			if (ifp->if_agrprivate) {
783 				ieee8023ad_marker_input(ifp, m);
784 				return;
785 			}
786 			break;
787 #endif
788 
789 		default:
790 			if (subtype == 0 || subtype > 10) {
791 				/* illegal value */
792 				m_freem(m);
793 				return;
794 			}
795 			/* unknown subtype */
796 			break;
797 		}
798 	}
799 	/* FALLTHROUGH */
800 	default:
801 		if (m->m_flags & M_PROMISC) {
802 			m_freem(m);
803 			return;
804 		}
805 	}
806 
807 	/* If the CRC is still on the packet, trim it off. */
808 	if (m->m_flags & M_HASFCS) {
809 		m_adj(m, -ETHER_CRC_LEN);
810 		m->m_flags &= ~M_HASFCS;
811 	}
812 
813 	if (etype > ETHERMTU + sizeof(struct ether_header)) {
814 		/* Strip off the Ethernet header. */
815 		m_adj(m, ehlen);
816 
817 		switch (etype) {
818 #ifdef INET
819 		case ETHERTYPE_IP:
820 #ifdef GATEWAY
821 			if (ipflow_fastforward(m))
822 				return;
823 #endif
824 			pktq = ip_pktq;
825 			break;
826 
827 		case ETHERTYPE_ARP:
828 			isr = NETISR_ARP;
829 			inq = &arpintrq;
830 			break;
831 
832 		case ETHERTYPE_REVARP:
833 			revarpinput(m);	/* XXX queue? */
834 			return;
835 #endif
836 
837 #ifdef INET6
838 		case ETHERTYPE_IPV6:
839 			if (__predict_false(!in6_present)) {
840 				m_freem(m);
841 				return;
842 			}
843 #ifdef GATEWAY
844 			if (ip6flow_fastforward(&m))
845 				return;
846 #endif
847 			pktq = ip6_pktq;
848 			break;
849 #endif
850 
851 #ifdef NETATALK
852 		case ETHERTYPE_ATALK:
853 			isr = NETISR_ATALK;
854 			inq = &atintrq1;
855 			break;
856 
857 		case ETHERTYPE_AARP:
858 			aarpinput(ifp, m); /* XXX queue? */
859 			return;
860 #endif
861 
862 #ifdef MPLS
863 		case ETHERTYPE_MPLS:
864 			isr = NETISR_MPLS;
865 			inq = &mplsintrq;
866 			break;
867 #endif
868 
869 		default:
870 			m_freem(m);
871 			return;
872 		}
873 	} else {
874 		KASSERT(ehlen == sizeof(*eh));
875 #if defined (LLC) || defined (NETATALK)
876 		if (m->m_len < sizeof(*eh) + sizeof(struct llc)) {
877 			goto dropanyway;
878 		}
879 		l = (struct llc *)(eh+1);
880 
881 		switch (l->llc_dsap) {
882 #ifdef NETATALK
883 		case LLC_SNAP_LSAP:
884 			switch (l->llc_control) {
885 			case LLC_UI:
886 				if (l->llc_ssap != LLC_SNAP_LSAP) {
887 					goto dropanyway;
888 				}
889 
890 				if (memcmp(&(l->llc_snap_org_code)[0],
891 				    at_org_code, sizeof(at_org_code)) == 0 &&
892 				    ntohs(l->llc_snap_ether_type) ==
893 				    ETHERTYPE_ATALK) {
894 					inq = &atintrq2;
895 					m_adj(m, sizeof(struct ether_header)
896 					    + sizeof(struct llc));
897 					isr = NETISR_ATALK;
898 					break;
899 				}
900 
901 				if (memcmp(&(l->llc_snap_org_code)[0],
902 				    aarp_org_code,
903 				    sizeof(aarp_org_code)) == 0 &&
904 				    ntohs(l->llc_snap_ether_type) ==
905 				    ETHERTYPE_AARP) {
906 					m_adj(m, sizeof(struct ether_header)
907 					    + sizeof(struct llc));
908 					aarpinput(ifp, m); /* XXX queue? */
909 					return;
910 				}
911 
912 			default:
913 				goto dropanyway;
914 			}
915 			break;
916 #endif
917 		dropanyway:
918 		default:
919 			m_freem(m);
920 			return;
921 		}
922 #else /* LLC || NETATALK */
923 		m_freem(m);
924 		return;
925 #endif /* LLC || NETATALK */
926 	}
927 
928 	if (__predict_true(pktq)) {
929 #ifdef NET_MPSAFE
930 		const u_int h = curcpu()->ci_index;
931 #else
932 		const uint32_t h = pktq_rps_hash(m);
933 #endif
934 		if (__predict_false(!pktq_enqueue(pktq, m, h))) {
935 			m_freem(m);
936 		}
937 		return;
938 	}
939 
940 	if (__predict_false(!inq)) {
941 		/* Should not happen. */
942 		m_freem(m);
943 		return;
944 	}
945 
946 	IFQ_LOCK(inq);
947 	if (IF_QFULL(inq)) {
948 		IF_DROP(inq);
949 		IFQ_UNLOCK(inq);
950 		m_freem(m);
951 	} else {
952 		IF_ENQUEUE(inq, m);
953 		IFQ_UNLOCK(inq);
954 		schednetisr(isr);
955 	}
956 }
957 
958 /*
959  * Convert Ethernet address to printable (loggable) representation.
960  */
961 char *
962 ether_sprintf(const u_char *ap)
963 {
964 	static char etherbuf[3 * ETHER_ADDR_LEN];
965 	return ether_snprintf(etherbuf, sizeof(etherbuf), ap);
966 }
967 
968 char *
969 ether_snprintf(char *buf, size_t len, const u_char *ap)
970 {
971 	char *cp = buf;
972 	size_t i;
973 
974 	for (i = 0; i < len / 3; i++) {
975 		*cp++ = hexdigits[*ap >> 4];
976 		*cp++ = hexdigits[*ap++ & 0xf];
977 		*cp++ = ':';
978 	}
979 	*--cp = '\0';
980 	return buf;
981 }
982 
983 /*
984  * Perform common duties while attaching to interface list
985  */
986 void
987 ether_ifattach(struct ifnet *ifp, const uint8_t *lla)
988 {
989 	struct ethercom *ec = (struct ethercom *)ifp;
990 
991 	ifp->if_type = IFT_ETHER;
992 	ifp->if_hdrlen = ETHER_HDR_LEN;
993 	ifp->if_dlt = DLT_EN10MB;
994 	ifp->if_mtu = ETHERMTU;
995 	ifp->if_output = ether_output;
996 	ifp->_if_input = ether_input;
997 	if (ifp->if_baudrate == 0)
998 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
999 
1000 	if (lla != NULL)
1001 		if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla));
1002 
1003 	LIST_INIT(&ec->ec_multiaddrs);
1004 	SIMPLEQ_INIT(&ec->ec_vids);
1005 	ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
1006 	ec->ec_flags = 0;
1007 	ifp->if_broadcastaddr = etherbroadcastaddr;
1008 	bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header));
1009 #ifdef MBUFTRACE
1010 	strlcpy(ec->ec_tx_mowner.mo_name, ifp->if_xname,
1011 	    sizeof(ec->ec_tx_mowner.mo_name));
1012 	strlcpy(ec->ec_tx_mowner.mo_descr, "tx",
1013 	    sizeof(ec->ec_tx_mowner.mo_descr));
1014 	strlcpy(ec->ec_rx_mowner.mo_name, ifp->if_xname,
1015 	    sizeof(ec->ec_rx_mowner.mo_name));
1016 	strlcpy(ec->ec_rx_mowner.mo_descr, "rx",
1017 	    sizeof(ec->ec_rx_mowner.mo_descr));
1018 	MOWNER_ATTACH(&ec->ec_tx_mowner);
1019 	MOWNER_ATTACH(&ec->ec_rx_mowner);
1020 	ifp->if_mowner = &ec->ec_tx_mowner;
1021 #endif
1022 }
1023 
1024 void
1025 ether_ifdetach(struct ifnet *ifp)
1026 {
1027 	struct ethercom *ec = (void *) ifp;
1028 	struct ether_multi *enm;
1029 
1030 	IFNET_ASSERT_UNLOCKED(ifp);
1031 	/*
1032 	 * Prevent further calls to ioctl (for example turning off
1033 	 * promiscuous mode from the bridge code), which eventually can
1034 	 * call if_init() which can cause panics because the interface
1035 	 * is in the process of being detached. Return device not configured
1036 	 * instead.
1037 	 */
1038 	ifp->if_ioctl = __FPTRCAST(int (*)(struct ifnet *, u_long, void *),
1039 	    enxio);
1040 
1041 #if NBRIDGE > 0
1042 	if (ifp->if_bridge)
1043 		bridge_ifdetach(ifp);
1044 #endif
1045 	bpf_detach(ifp);
1046 #if NVLAN > 0
1047 	if (ec->ec_nvlans)
1048 		vlan_ifdetach(ifp);
1049 #endif
1050 
1051 	ETHER_LOCK(ec);
1052 	KASSERT(ec->ec_nvlans == 0);
1053 	while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) {
1054 		LIST_REMOVE(enm, enm_list);
1055 		kmem_free(enm, sizeof(*enm));
1056 		ec->ec_multicnt--;
1057 	}
1058 	ETHER_UNLOCK(ec);
1059 
1060 	mutex_obj_free(ec->ec_lock);
1061 	ec->ec_lock = NULL;
1062 
1063 	ifp->if_mowner = NULL;
1064 	MOWNER_DETACH(&ec->ec_rx_mowner);
1065 	MOWNER_DETACH(&ec->ec_tx_mowner);
1066 }
1067 
1068 #if 0
1069 /*
1070  * This is for reference.  We have a table-driven version
1071  * of the little-endian crc32 generator, which is faster
1072  * than the double-loop.
1073  */
1074 uint32_t
1075 ether_crc32_le(const uint8_t *buf, size_t len)
1076 {
1077 	uint32_t c, crc, carry;
1078 	size_t i, j;
1079 
1080 	crc = 0xffffffffU;	/* initial value */
1081 
1082 	for (i = 0; i < len; i++) {
1083 		c = buf[i];
1084 		for (j = 0; j < 8; j++) {
1085 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1086 			crc >>= 1;
1087 			c >>= 1;
1088 			if (carry)
1089 				crc = (crc ^ ETHER_CRC_POLY_LE);
1090 		}
1091 	}
1092 
1093 	return (crc);
1094 }
1095 #else
1096 uint32_t
1097 ether_crc32_le(const uint8_t *buf, size_t len)
1098 {
1099 	static const uint32_t crctab[] = {
1100 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1101 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1102 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1103 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1104 	};
1105 	uint32_t crc;
1106 	size_t i;
1107 
1108 	crc = 0xffffffffU;	/* initial value */
1109 
1110 	for (i = 0; i < len; i++) {
1111 		crc ^= buf[i];
1112 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1113 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1114 	}
1115 
1116 	return (crc);
1117 }
1118 #endif
1119 
1120 uint32_t
1121 ether_crc32_be(const uint8_t *buf, size_t len)
1122 {
1123 	uint32_t c, crc, carry;
1124 	size_t i, j;
1125 
1126 	crc = 0xffffffffU;	/* initial value */
1127 
1128 	for (i = 0; i < len; i++) {
1129 		c = buf[i];
1130 		for (j = 0; j < 8; j++) {
1131 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1132 			crc <<= 1;
1133 			c >>= 1;
1134 			if (carry)
1135 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1136 		}
1137 	}
1138 
1139 	return (crc);
1140 }
1141 
1142 #ifdef INET
1143 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] =
1144     { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 };
1145 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] =
1146     { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff };
1147 #endif
1148 #ifdef INET6
1149 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] =
1150     { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 };
1151 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] =
1152     { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff };
1153 #endif
1154 
1155 /*
1156  * ether_aton implementation, not using a static buffer.
1157  */
1158 int
1159 ether_aton_r(u_char *dest, size_t len, const char *str)
1160 {
1161 	const u_char *cp = (const void *)str;
1162 	u_char *ep;
1163 
1164 #define atox(c)	(((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10))
1165 
1166 	if (len < ETHER_ADDR_LEN)
1167 		return ENOSPC;
1168 
1169 	ep = dest + ETHER_ADDR_LEN;
1170 
1171 	while (*cp) {
1172 		if (!isxdigit(*cp))
1173 			return EINVAL;
1174 
1175 		*dest = atox(*cp);
1176 		cp++;
1177 		if (isxdigit(*cp)) {
1178 			*dest = (*dest << 4) | atox(*cp);
1179 			cp++;
1180 		}
1181 		dest++;
1182 
1183 		if (dest == ep)
1184 			return (*cp == '\0') ? 0 : ENAMETOOLONG;
1185 
1186 		switch (*cp) {
1187 		case ':':
1188 		case '-':
1189 		case '.':
1190 			cp++;
1191 			break;
1192 		}
1193 	}
1194 	return ENOBUFS;
1195 }
1196 
1197 /*
1198  * Convert a sockaddr into an Ethernet address or range of Ethernet
1199  * addresses.
1200  */
1201 int
1202 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN],
1203     uint8_t addrhi[ETHER_ADDR_LEN])
1204 {
1205 #ifdef INET
1206 	const struct sockaddr_in *sin;
1207 #endif
1208 #ifdef INET6
1209 	const struct sockaddr_in6 *sin6;
1210 #endif
1211 
1212 	switch (sa->sa_family) {
1213 
1214 	case AF_UNSPEC:
1215 		memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN);
1216 		memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1217 		break;
1218 
1219 #ifdef INET
1220 	case AF_INET:
1221 		sin = satocsin(sa);
1222 		if (sin->sin_addr.s_addr == INADDR_ANY) {
1223 			/*
1224 			 * An IP address of INADDR_ANY means listen to
1225 			 * or stop listening to all of the Ethernet
1226 			 * multicast addresses used for IP.
1227 			 * (This is for the sake of IP multicast routers.)
1228 			 */
1229 			memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN);
1230 			memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN);
1231 		} else {
1232 			ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo);
1233 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1234 		}
1235 		break;
1236 #endif
1237 #ifdef INET6
1238 	case AF_INET6:
1239 		sin6 = satocsin6(sa);
1240 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1241 			/*
1242 			 * An IP6 address of 0 means listen to or stop
1243 			 * listening to all of the Ethernet multicast
1244 			 * address used for IP6.
1245 			 * (This is used for multicast routers.)
1246 			 */
1247 			memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN);
1248 			memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN);
1249 		} else {
1250 			ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo);
1251 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1252 		}
1253 		break;
1254 #endif
1255 
1256 	default:
1257 		return EAFNOSUPPORT;
1258 	}
1259 	return 0;
1260 }
1261 
1262 /*
1263  * Add an Ethernet multicast address or range of addresses to the list for a
1264  * given interface.
1265  */
1266 int
1267 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec)
1268 {
1269 	struct ether_multi *enm, *_enm;
1270 	u_char addrlo[ETHER_ADDR_LEN];
1271 	u_char addrhi[ETHER_ADDR_LEN];
1272 	int error = 0;
1273 
1274 	/* Allocate out of lock */
1275 	enm = kmem_alloc(sizeof(*enm), KM_SLEEP);
1276 
1277 	ETHER_LOCK(ec);
1278 	error = ether_multiaddr(sa, addrlo, addrhi);
1279 	if (error != 0)
1280 		goto out;
1281 
1282 	/*
1283 	 * Verify that we have valid Ethernet multicast addresses.
1284 	 */
1285 	if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) {
1286 		error = EINVAL;
1287 		goto out;
1288 	}
1289 
1290 	/*
1291 	 * See if the address range is already in the list.
1292 	 */
1293 	_enm = ether_lookup_multi(addrlo, addrhi, ec);
1294 	if (_enm != NULL) {
1295 		/*
1296 		 * Found it; just increment the reference count.
1297 		 */
1298 		++_enm->enm_refcount;
1299 		error = 0;
1300 		goto out;
1301 	}
1302 
1303 	/*
1304 	 * Link a new multicast record into the interface's multicast list.
1305 	 */
1306 	memcpy(enm->enm_addrlo, addrlo, ETHER_ADDR_LEN);
1307 	memcpy(enm->enm_addrhi, addrhi, ETHER_ADDR_LEN);
1308 	enm->enm_refcount = 1;
1309 	LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list);
1310 	ec->ec_multicnt++;
1311 
1312 	/*
1313 	 * Return ENETRESET to inform the driver that the list has changed
1314 	 * and its reception filter should be adjusted accordingly.
1315 	 */
1316 	error = ENETRESET;
1317 	enm = NULL;
1318 
1319 out:
1320 	ETHER_UNLOCK(ec);
1321 	if (enm != NULL)
1322 		kmem_free(enm, sizeof(*enm));
1323 	return error;
1324 }
1325 
1326 /*
1327  * Delete a multicast address record.
1328  */
1329 int
1330 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec)
1331 {
1332 	struct ether_multi *enm;
1333 	u_char addrlo[ETHER_ADDR_LEN];
1334 	u_char addrhi[ETHER_ADDR_LEN];
1335 	int error;
1336 
1337 	ETHER_LOCK(ec);
1338 	error = ether_multiaddr(sa, addrlo, addrhi);
1339 	if (error != 0)
1340 		goto error;
1341 
1342 	/*
1343 	 * Look up the address in our list.
1344 	 */
1345 	enm = ether_lookup_multi(addrlo, addrhi, ec);
1346 	if (enm == NULL) {
1347 		error = ENXIO;
1348 		goto error;
1349 	}
1350 	if (--enm->enm_refcount != 0) {
1351 		/*
1352 		 * Still some claims to this record.
1353 		 */
1354 		error = 0;
1355 		goto error;
1356 	}
1357 
1358 	/*
1359 	 * No remaining claims to this record; unlink and free it.
1360 	 */
1361 	LIST_REMOVE(enm, enm_list);
1362 	ec->ec_multicnt--;
1363 	ETHER_UNLOCK(ec);
1364 	kmem_free(enm, sizeof(*enm));
1365 
1366 	/*
1367 	 * Return ENETRESET to inform the driver that the list has changed
1368 	 * and its reception filter should be adjusted accordingly.
1369 	 */
1370 	return ENETRESET;
1371 
1372 error:
1373 	ETHER_UNLOCK(ec);
1374 	return error;
1375 }
1376 
1377 void
1378 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb)
1379 {
1380 	ec->ec_ifflags_cb = cb;
1381 }
1382 
1383 void
1384 ether_set_vlan_cb(struct ethercom *ec, ether_vlancb_t cb)
1385 {
1386 
1387 	ec->ec_vlan_cb = cb;
1388 }
1389 
1390 static int
1391 ether_ioctl_reinit(struct ethercom *ec)
1392 {
1393 	struct ifnet *ifp = &ec->ec_if;
1394 	int error;
1395 
1396 	switch (ifp->if_flags & (IFF_UP | IFF_RUNNING)) {
1397 	case IFF_RUNNING:
1398 		/*
1399 		 * If interface is marked down and it is running,
1400 		 * then stop and disable it.
1401 		 */
1402 		(*ifp->if_stop)(ifp, 1);
1403 		break;
1404 	case IFF_UP:
1405 		/*
1406 		 * If interface is marked up and it is stopped, then
1407 		 * start it.
1408 		 */
1409 		return (*ifp->if_init)(ifp);
1410 	case IFF_UP | IFF_RUNNING:
1411 		error = 0;
1412 		if (ec->ec_ifflags_cb != NULL) {
1413 			error = (*ec->ec_ifflags_cb)(ec);
1414 			if (error == ENETRESET) {
1415 				/*
1416 				 * Reset the interface to pick up
1417 				 * changes in any other flags that
1418 				 * affect the hardware state.
1419 				 */
1420 				return (*ifp->if_init)(ifp);
1421 			}
1422 		} else
1423 			error = (*ifp->if_init)(ifp);
1424 		return error;
1425 	case 0:
1426 		break;
1427 	}
1428 
1429 	return 0;
1430 }
1431 
1432 /*
1433  * Common ioctls for Ethernet interfaces.  Note, we must be
1434  * called at splnet().
1435  */
1436 int
1437 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data)
1438 {
1439 	struct ethercom *ec = (void *)ifp;
1440 	struct eccapreq *eccr;
1441 	struct ifreq *ifr = (struct ifreq *)data;
1442 	struct if_laddrreq *iflr = data;
1443 	const struct sockaddr_dl *sdl;
1444 	static const uint8_t zero[ETHER_ADDR_LEN];
1445 	int error;
1446 
1447 	switch (cmd) {
1448 	case SIOCINITIFADDR:
1449 	    {
1450 		struct ifaddr *ifa = (struct ifaddr *)data;
1451 		if (ifa->ifa_addr->sa_family != AF_LINK
1452 		    && (ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
1453 		       (IFF_UP | IFF_RUNNING)) {
1454 			ifp->if_flags |= IFF_UP;
1455 			if ((error = (*ifp->if_init)(ifp)) != 0)
1456 				return error;
1457 		}
1458 #ifdef INET
1459 		if (ifa->ifa_addr->sa_family == AF_INET)
1460 			arp_ifinit(ifp, ifa);
1461 #endif
1462 		return 0;
1463 	    }
1464 
1465 	case SIOCSIFMTU:
1466 	    {
1467 		int maxmtu;
1468 
1469 		if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU)
1470 			maxmtu = ETHERMTU_JUMBO;
1471 		else
1472 			maxmtu = ETHERMTU;
1473 
1474 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu)
1475 			return EINVAL;
1476 		else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET)
1477 			return error;
1478 		else if (ifp->if_flags & IFF_UP) {
1479 			/* Make sure the device notices the MTU change. */
1480 			return (*ifp->if_init)(ifp);
1481 		} else
1482 			return 0;
1483 	    }
1484 
1485 	case SIOCSIFFLAGS:
1486 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
1487 			return error;
1488 		return ether_ioctl_reinit(ec);
1489 	case SIOCGIFFLAGS:
1490 		error = ifioctl_common(ifp, cmd, data);
1491 		if (error == 0) {
1492 			/* Set IFF_ALLMULTI for backcompat */
1493 			ifr->ifr_flags |= (ec->ec_flags & ETHER_F_ALLMULTI) ?
1494 			    IFF_ALLMULTI : 0;
1495 		}
1496 		return error;
1497 	case SIOCGETHERCAP:
1498 		eccr = (struct eccapreq *)data;
1499 		eccr->eccr_capabilities = ec->ec_capabilities;
1500 		eccr->eccr_capenable = ec->ec_capenable;
1501 		return 0;
1502 	case SIOCSETHERCAP:
1503 		eccr = (struct eccapreq *)data;
1504 		if ((eccr->eccr_capenable & ~ec->ec_capabilities) != 0)
1505 			return EINVAL;
1506 		if (eccr->eccr_capenable == ec->ec_capenable)
1507 			return 0;
1508 #if 0 /* notyet */
1509 		ec->ec_capenable = (ec->ec_capenable & ETHERCAP_CANTCHANGE)
1510 		    | (eccr->eccr_capenable & ~ETHERCAP_CANTCHANGE);
1511 #else
1512 		ec->ec_capenable = eccr->eccr_capenable;
1513 #endif
1514 		return ether_ioctl_reinit(ec);
1515 	case SIOCADDMULTI:
1516 		return ether_addmulti(ifreq_getaddr(cmd, ifr), ec);
1517 	case SIOCDELMULTI:
1518 		return ether_delmulti(ifreq_getaddr(cmd, ifr), ec);
1519 	case SIOCSIFMEDIA:
1520 	case SIOCGIFMEDIA:
1521 		if (ec->ec_mii != NULL)
1522 			return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media,
1523 			    cmd);
1524 		else if (ec->ec_ifmedia != NULL)
1525 			return ifmedia_ioctl(ifp, ifr, ec->ec_ifmedia, cmd);
1526 		else
1527 			return ENOTTY;
1528 		break;
1529 	case SIOCALIFADDR:
1530 		sdl = satocsdl(sstocsa(&iflr->addr));
1531 		if (sdl->sdl_family != AF_LINK)
1532 			;
1533 		else if (ETHER_IS_MULTICAST(CLLADDR(sdl)))
1534 			return EINVAL;
1535 		else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0)
1536 			return EINVAL;
1537 		/*FALLTHROUGH*/
1538 	default:
1539 		return ifioctl_common(ifp, cmd, data);
1540 	}
1541 	return 0;
1542 }
1543 
1544 /*
1545  * Enable/disable passing VLAN packets if the parent interface supports it.
1546  * Return:
1547  * 	 0: Ok
1548  *	-1: Parent interface does not support vlans
1549  *	>0: Error
1550  */
1551 int
1552 ether_enable_vlan_mtu(struct ifnet *ifp)
1553 {
1554 	int error;
1555 	struct ethercom *ec = (void *)ifp;
1556 
1557 	/* Parent does not support VLAN's */
1558 	if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0)
1559 		return -1;
1560 
1561 	/*
1562 	 * Parent supports the VLAN_MTU capability,
1563 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames;
1564 	 * enable it.
1565 	 */
1566 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1567 
1568 	/* Interface is down, defer for later */
1569 	if ((ifp->if_flags & IFF_UP) == 0)
1570 		return 0;
1571 
1572 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1573 		return 0;
1574 
1575 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1576 	return error;
1577 }
1578 
1579 int
1580 ether_disable_vlan_mtu(struct ifnet *ifp)
1581 {
1582 	int error;
1583 	struct ethercom *ec = (void *)ifp;
1584 
1585 	/* We still have VLAN's, defer for later */
1586 	if (ec->ec_nvlans != 0)
1587 		return 0;
1588 
1589 	/* Parent does not support VLAB's, nothing to do. */
1590 	if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0)
1591 		return -1;
1592 
1593 	/*
1594 	 * Disable Tx/Rx of VLAN-sized frames.
1595 	 */
1596 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1597 
1598 	/* Interface is down, defer for later */
1599 	if ((ifp->if_flags & IFF_UP) == 0)
1600 		return 0;
1601 
1602 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1603 		return 0;
1604 
1605 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1606 	return error;
1607 }
1608 
1609 static int
1610 ether_multicast_sysctl(SYSCTLFN_ARGS)
1611 {
1612 	struct ether_multi *enm;
1613 	struct ifnet *ifp;
1614 	struct ethercom *ec;
1615 	int error = 0;
1616 	size_t written;
1617 	struct psref psref;
1618 	int bound;
1619 	unsigned int multicnt;
1620 	struct ether_multi_sysctl *addrs;
1621 	int i;
1622 
1623 	if (namelen != 1)
1624 		return EINVAL;
1625 
1626 	bound = curlwp_bind();
1627 	ifp = if_get_byindex(name[0], &psref);
1628 	if (ifp == NULL) {
1629 		error = ENODEV;
1630 		goto out;
1631 	}
1632 	if (ifp->if_type != IFT_ETHER) {
1633 		if_put(ifp, &psref);
1634 		*oldlenp = 0;
1635 		goto out;
1636 	}
1637 	ec = (struct ethercom *)ifp;
1638 
1639 	if (oldp == NULL) {
1640 		if_put(ifp, &psref);
1641 		*oldlenp = ec->ec_multicnt * sizeof(*addrs);
1642 		goto out;
1643 	}
1644 
1645 	/*
1646 	 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which
1647 	 * is sleepable, while holding it. Copy data to a local buffer first
1648 	 * with the lock taken and then call sysctl_copyout without holding it.
1649 	 */
1650 retry:
1651 	multicnt = ec->ec_multicnt;
1652 
1653 	if (multicnt == 0) {
1654 		if_put(ifp, &psref);
1655 		*oldlenp = 0;
1656 		goto out;
1657 	}
1658 
1659 	addrs = kmem_zalloc(sizeof(*addrs) * multicnt, KM_SLEEP);
1660 
1661 	ETHER_LOCK(ec);
1662 	if (multicnt != ec->ec_multicnt) {
1663 		/* The number of multicast addresses has changed */
1664 		ETHER_UNLOCK(ec);
1665 		kmem_free(addrs, sizeof(*addrs) * multicnt);
1666 		goto retry;
1667 	}
1668 
1669 	i = 0;
1670 	LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) {
1671 		struct ether_multi_sysctl *addr = &addrs[i];
1672 		addr->enm_refcount = enm->enm_refcount;
1673 		memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN);
1674 		memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN);
1675 		i++;
1676 	}
1677 	ETHER_UNLOCK(ec);
1678 
1679 	error = 0;
1680 	written = 0;
1681 	for (i = 0; i < multicnt; i++) {
1682 		struct ether_multi_sysctl *addr = &addrs[i];
1683 
1684 		if (written + sizeof(*addr) > *oldlenp)
1685 			break;
1686 		error = sysctl_copyout(l, addr, oldp, sizeof(*addr));
1687 		if (error)
1688 			break;
1689 		written += sizeof(*addr);
1690 		oldp = (char *)oldp + sizeof(*addr);
1691 	}
1692 	kmem_free(addrs, sizeof(*addrs) * multicnt);
1693 
1694 	if_put(ifp, &psref);
1695 
1696 	*oldlenp = written;
1697 out:
1698 	curlwp_bindx(bound);
1699 	return error;
1700 }
1701 
1702 static void
1703 ether_sysctl_setup(struct sysctllog **clog)
1704 {
1705 	const struct sysctlnode *rnode = NULL;
1706 
1707 	sysctl_createv(clog, 0, NULL, &rnode,
1708 		       CTLFLAG_PERMANENT,
1709 		       CTLTYPE_NODE, "ether",
1710 		       SYSCTL_DESCR("Ethernet-specific information"),
1711 		       NULL, 0, NULL, 0,
1712 		       CTL_NET, CTL_CREATE, CTL_EOL);
1713 
1714 	sysctl_createv(clog, 0, &rnode, NULL,
1715 		       CTLFLAG_PERMANENT,
1716 		       CTLTYPE_NODE, "multicast",
1717 		       SYSCTL_DESCR("multicast addresses"),
1718 		       ether_multicast_sysctl, 0, NULL, 0,
1719 		       CTL_CREATE, CTL_EOL);
1720 }
1721 
1722 void
1723 etherinit(void)
1724 {
1725 
1726 #ifdef DIAGNOSTIC
1727 	mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET);
1728 #endif
1729 	ether_sysctl_setup(NULL);
1730 }
1731