xref: /netbsd-src/sys/net/if_ethersubr.c (revision cc576e1d8e4f4078fd4e81238abca9fca216f6ec)
1 /*	$NetBSD: if_ethersubr.c,v 1.236 2017/01/24 18:37:20 maxv 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 1.236 2017/01/24 18:37:20 maxv 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/malloc.h>
84 #include <sys/mbuf.h>
85 #include <sys/mutex.h>
86 #include <sys/ioctl.h>
87 #include <sys/errno.h>
88 #include <sys/device.h>
89 #include <sys/rnd.h>
90 #include <sys/rndsource.h>
91 #include <sys/cpu.h>
92 #include <sys/kmem.h>
93 
94 #include <net/if.h>
95 #include <net/netisr.h>
96 #include <net/route.h>
97 #include <net/if_llc.h>
98 #include <net/if_dl.h>
99 #include <net/if_types.h>
100 #include <net/pktqueue.h>
101 
102 #include <net/if_media.h>
103 #include <dev/mii/mii.h>
104 #include <dev/mii/miivar.h>
105 
106 #if NARP == 0
107 /*
108  * XXX there should really be a way to issue this warning from within config(8)
109  */
110 #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.
111 #endif
112 
113 #include <net/bpf.h>
114 
115 #include <net/if_ether.h>
116 #include <net/if_vlanvar.h>
117 
118 #if NPPPOE > 0
119 #include <net/if_pppoe.h>
120 #endif
121 
122 #if NAGR > 0
123 #include <net/agr/ieee8023_slowprotocols.h>	/* XXX */
124 #include <net/agr/ieee8023ad.h>
125 #include <net/agr/if_agrvar.h>
126 #endif
127 
128 #if NBRIDGE > 0
129 #include <net/if_bridgevar.h>
130 #endif
131 
132 #include <netinet/in.h>
133 #ifdef INET
134 #include <netinet/in_var.h>
135 #endif
136 #include <netinet/if_inarp.h>
137 
138 #ifdef INET6
139 #ifndef INET
140 #include <netinet/in.h>
141 #endif
142 #include <netinet6/in6_var.h>
143 #include <netinet6/nd6.h>
144 #endif
145 
146 
147 #include "carp.h"
148 #if NCARP > 0
149 #include <netinet/ip_carp.h>
150 #endif
151 
152 #ifdef NETATALK
153 #include <netatalk/at.h>
154 #include <netatalk/at_var.h>
155 #include <netatalk/at_extern.h>
156 
157 #define llc_snap_org_code llc_un.type_snap.org_code
158 #define llc_snap_ether_type llc_un.type_snap.ether_type
159 
160 extern u_char	at_org_code[3];
161 extern u_char	aarp_org_code[3];
162 #endif /* NETATALK */
163 
164 #ifdef MPLS
165 #include <netmpls/mpls.h>
166 #include <netmpls/mpls_var.h>
167 #endif
168 
169 static struct timeval bigpktppslim_last;
170 static int bigpktppslim = 2;	/* XXX */
171 static int bigpktpps_count;
172 static kmutex_t bigpktpps_lock __cacheline_aligned;
173 
174 
175 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] =
176     { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
177 const uint8_t ethermulticastaddr_slowprotocols[ETHER_ADDR_LEN] =
178     { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x02 };
179 #define senderr(e) { error = (e); goto bad;}
180 
181 static	int ether_output(struct ifnet *, struct mbuf *,
182 	    const struct sockaddr *, const struct rtentry *);
183 
184 /*
185  * Ethernet output routine.
186  * Encapsulate a packet of type family for the local net.
187  * Assumes that ifp is actually pointer to ethercom structure.
188  */
189 static int
190 ether_output(struct ifnet * const ifp0, struct mbuf * const m0,
191 	const struct sockaddr * const dst,
192 	const struct rtentry *rt)
193 {
194 	uint16_t etype = 0;
195 	int error = 0, hdrcmplt = 0;
196  	uint8_t esrc[6], edst[6];
197 	struct mbuf *m = m0;
198 	struct mbuf *mcopy = NULL;
199 	struct ether_header *eh;
200 	struct ifnet *ifp = ifp0;
201 #ifdef INET
202 	struct arphdr *ah;
203 #endif /* INET */
204 #ifdef NETATALK
205 	struct at_ifaddr *aa;
206 #endif /* NETATALK */
207 
208 	/*
209 	 * some paths such as carp_output() call ethr_output() with "ifp"
210 	 * argument as other than ether ifnet.
211 	 */
212 	KASSERT(ifp->if_output != ether_output
213 	    || ifp->if_extflags & IFEF_OUTPUT_MPSAFE);
214 
215 #ifdef MBUFTRACE
216 	m_claimm(m, ifp->if_mowner);
217 #endif
218 
219 #if NCARP > 0
220 	if (ifp->if_type == IFT_CARP) {
221 		struct ifaddr *ifa;
222 		int s = pserialize_read_enter();
223 
224 		/* loop back if this is going to the carp interface */
225 		if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP &&
226 		    (ifa = ifa_ifwithaddr(dst)) != NULL) {
227 			if (ifa->ifa_ifp == ifp0) {
228 				pserialize_read_exit(s);
229 				return looutput(ifp0, m, dst, rt);
230 			}
231 		}
232 		pserialize_read_exit(s);
233 
234 		ifp = ifp->if_carpdev;
235 		/* ac = (struct arpcom *)ifp; */
236 
237 		if ((ifp0->if_flags & (IFF_UP|IFF_RUNNING)) !=
238 		    (IFF_UP|IFF_RUNNING))
239 			senderr(ENETDOWN);
240 	}
241 #endif /* NCARP > 0 */
242 
243 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
244 		senderr(ENETDOWN);
245 
246 	switch (dst->sa_family) {
247 
248 #ifdef INET
249 	case AF_INET:
250 #ifndef NET_MPSAFE
251 		KERNEL_LOCK(1, NULL);
252 #endif
253 		if (m->m_flags & M_BCAST)
254 			(void)memcpy(edst, etherbroadcastaddr, sizeof(edst));
255 		else if (m->m_flags & M_MCAST)
256 			ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst);
257 		else if ((error = arpresolve(ifp, rt, m, dst, edst,
258 		    sizeof(edst))) != 0) {
259 #ifndef NET_MPSAFE
260 			KERNEL_UNLOCK_ONE(NULL);
261 #endif
262 			return error == EWOULDBLOCK ? 0 : error;
263 		}
264 #ifndef NET_MPSAFE
265 		KERNEL_UNLOCK_ONE(NULL);
266 #endif
267 		/* If broadcasting on a simplex interface, loopback a copy */
268 		if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
269 			mcopy = m_copy(m, 0, (int)M_COPYALL);
270 		etype = htons(ETHERTYPE_IP);
271 		break;
272 
273 	case AF_ARP:
274 		ah = mtod(m, struct arphdr *);
275 		if (m->m_flags & M_BCAST)
276 			(void)memcpy(edst, etherbroadcastaddr, sizeof(edst));
277 		else {
278 			void *tha = ar_tha(ah);
279 
280 			if (tha == NULL) {
281 				/* fake with ARPHDR_IEEE1394 */
282 				m_freem(m);
283 				return 0;
284 			}
285 			memcpy(edst, tha, sizeof(edst));
286 		}
287 
288 		ah->ar_hrd = htons(ARPHRD_ETHER);
289 
290 		switch (ntohs(ah->ar_op)) {
291 		case ARPOP_REVREQUEST:
292 		case ARPOP_REVREPLY:
293 			etype = htons(ETHERTYPE_REVARP);
294 			break;
295 
296 		case ARPOP_REQUEST:
297 		case ARPOP_REPLY:
298 		default:
299 			etype = htons(ETHERTYPE_ARP);
300 		}
301 
302 		break;
303 #endif
304 #ifdef INET6
305 	case AF_INET6:
306 		if (!nd6_storelladdr(ifp, rt, m, dst, edst, sizeof(edst))){
307 			/* something bad happened */
308 			return (0);
309 		}
310 		etype = htons(ETHERTYPE_IPV6);
311 		break;
312 #endif
313 #ifdef NETATALK
314     case AF_APPLETALK: {
315 		struct ifaddr *ifa;
316 		int s;
317 
318 		KERNEL_LOCK(1, NULL);
319 		if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) {
320 #ifdef NETATALKDEBUG
321 			printf("aarpresolv failed\n");
322 #endif /* NETATALKDEBUG */
323 			KERNEL_UNLOCK_ONE(NULL);
324 			return (0);
325 		}
326 		/*
327 		 * ifaddr is the first thing in at_ifaddr
328 		 */
329 		s = pserialize_read_enter();
330 		ifa = at_ifawithnet((const struct sockaddr_at *)dst, ifp);
331 		if (ifa == NULL) {
332 			pserialize_read_exit(s);
333 			KERNEL_UNLOCK_ONE(NULL);
334 			goto bad;
335 		}
336 		aa = (struct at_ifaddr *)ifa;
337 
338 		/*
339 		 * In the phase 2 case, we need to prepend an mbuf for the
340 		 * llc header.  Since we must preserve the value of m,
341 		 * which is passed to us by value, we m_copy() the first
342 		 * mbuf, and use it for our llc header.
343 		 */
344 		if (aa->aa_flags & AFA_PHASE2) {
345 			struct llc llc;
346 
347 			M_PREPEND(m, sizeof(struct llc), M_DONTWAIT);
348 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
349 			llc.llc_control = LLC_UI;
350 			memcpy(llc.llc_snap_org_code, at_org_code,
351 			    sizeof(llc.llc_snap_org_code));
352 			llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK);
353 			memcpy(mtod(m, void *), &llc, sizeof(struct llc));
354 		} else {
355 			etype = htons(ETHERTYPE_ATALK);
356 		}
357 		pserialize_read_exit(s);
358 		KERNEL_UNLOCK_ONE(NULL);
359 		break;
360 	    }
361 #endif /* NETATALK */
362 	case pseudo_AF_HDRCMPLT:
363 		hdrcmplt = 1;
364 		memcpy(esrc,
365 		    ((const struct ether_header *)dst->sa_data)->ether_shost,
366 		    sizeof(esrc));
367 		/* FALLTHROUGH */
368 
369 	case AF_UNSPEC:
370  		memcpy(edst,
371 		    ((const struct ether_header *)dst->sa_data)->ether_dhost,
372 		    sizeof(edst));
373 		/* AF_UNSPEC doesn't swap the byte order of the ether_type. */
374 		etype = ((const struct ether_header *)dst->sa_data)->ether_type;
375 		break;
376 
377 	default:
378 		printf("%s: can't handle af%d\n", ifp->if_xname,
379 			dst->sa_family);
380 		senderr(EAFNOSUPPORT);
381 	}
382 
383 #ifdef MPLS
384 	KERNEL_LOCK(1, NULL);
385 	{
386 		struct m_tag *mtag;
387 		mtag = m_tag_find(m, PACKET_TAG_MPLS, NULL);
388 		if (mtag != NULL) {
389 			/* Having the tag itself indicates it's MPLS */
390 			etype = htons(ETHERTYPE_MPLS);
391 			m_tag_delete(m, mtag);
392 		}
393 	}
394 	KERNEL_UNLOCK_ONE(NULL);
395 #endif
396 
397 	if (mcopy)
398 		(void)looutput(ifp, mcopy, dst, rt);
399 
400 	/* If no ether type is set, this must be a 802.2 formatted packet.
401 	 */
402 	if (etype == 0)
403 		etype = htons(m->m_pkthdr.len);
404 	/*
405 	 * Add local net header.  If no space in first mbuf,
406 	 * allocate another.
407 	 */
408 	M_PREPEND(m, sizeof (struct ether_header), M_DONTWAIT);
409 	if (m == 0)
410 		senderr(ENOBUFS);
411 	eh = mtod(m, struct ether_header *);
412 	/* Note: etype is already in network byte order. */
413 	(void)memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
414  	memcpy(eh->ether_dhost, edst, sizeof(edst));
415 	if (hdrcmplt)
416 		memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost));
417 	else
418 	 	memcpy(eh->ether_shost, CLLADDR(ifp->if_sadl),
419 		    sizeof(eh->ether_shost));
420 
421 #if NCARP > 0
422 	if (ifp0 != ifp && ifp0->if_type == IFT_CARP) {
423 	 	memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl),
424 		    sizeof(eh->ether_shost));
425 	}
426 #endif /* NCARP > 0 */
427 
428 	if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0)
429 		return (error);
430 	if (m == NULL)
431 		return (0);
432 
433 #if NBRIDGE > 0
434 	/*
435 	 * Bridges require special output handling.
436 	 */
437 	if (ifp->if_bridge)
438 		return (bridge_output(ifp, m, NULL, NULL));
439 #endif
440 
441 #if NCARP > 0
442 	if (ifp != ifp0)
443 		ifp0->if_obytes += m->m_pkthdr.len + ETHER_HDR_LEN;
444 #endif /* NCARP > 0 */
445 
446 #ifdef ALTQ
447 	KERNEL_LOCK(1, NULL);
448 	/*
449 	 * If ALTQ is enabled on the parent interface, do
450 	 * classification; the queueing discipline might not
451 	 * require classification, but might require the
452 	 * address family/header pointer in the pktattr.
453 	 */
454 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
455 		altq_etherclassify(&ifp->if_snd, m);
456 	KERNEL_UNLOCK_ONE(NULL);
457 #endif
458 	return ifq_enqueue(ifp, m);
459 
460 bad:
461 	if (m)
462 		m_freem(m);
463 	return (error);
464 }
465 
466 #ifdef ALTQ
467 /*
468  * This routine is a slight hack to allow a packet to be classified
469  * if the Ethernet headers are present.  It will go away when ALTQ's
470  * classification engine understands link headers.
471  */
472 void
473 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m)
474 {
475 	struct ether_header *eh;
476 	uint16_t ether_type;
477 	int hlen, af, hdrsize;
478 	void *hdr;
479 
480 	hlen = ETHER_HDR_LEN;
481 	eh = mtod(m, struct ether_header *);
482 
483 	ether_type = htons(eh->ether_type);
484 
485 	if (ether_type < ETHERMTU) {
486 		/* LLC/SNAP */
487 		struct llc *llc = (struct llc *)(eh + 1);
488 		hlen += 8;
489 
490 		if (m->m_len < hlen ||
491 		    llc->llc_dsap != LLC_SNAP_LSAP ||
492 		    llc->llc_ssap != LLC_SNAP_LSAP ||
493 		    llc->llc_control != LLC_UI) {
494 			/* Not SNAP. */
495 			goto bad;
496 		}
497 
498 		ether_type = htons(llc->llc_un.type_snap.ether_type);
499 	}
500 
501 	switch (ether_type) {
502 	case ETHERTYPE_IP:
503 		af = AF_INET;
504 		hdrsize = 20;		/* sizeof(struct ip) */
505 		break;
506 
507 	case ETHERTYPE_IPV6:
508 		af = AF_INET6;
509 		hdrsize = 40;		/* sizeof(struct ip6_hdr) */
510 		break;
511 
512 	default:
513 		af = AF_UNSPEC;
514 		hdrsize = 0;
515 		break;
516 	}
517 
518 	while (m->m_len <= hlen) {
519 		hlen -= m->m_len;
520 		m = m->m_next;
521 	}
522 	if (m->m_len < (hlen + hdrsize)) {
523 		/*
524 		 * protocol header not in a single mbuf.
525 		 * We can't cope with this situation right
526 		 * now (but it shouldn't ever happen, really, anyhow).
527 		 */
528 #ifdef DEBUG
529 		printf("altq_etherclassify: headers span multiple mbufs: "
530 		    "%d < %d\n", m->m_len, (hlen + hdrsize));
531 #endif
532 		goto bad;
533 	}
534 
535 	m->m_data += hlen;
536 	m->m_len -= hlen;
537 
538 	hdr = mtod(m, void *);
539 
540 	if (ALTQ_NEEDS_CLASSIFY(ifq))
541 		m->m_pkthdr.pattr_class =
542 		    (*ifq->altq_classify)(ifq->altq_clfier, m, af);
543 	m->m_pkthdr.pattr_af = af;
544 	m->m_pkthdr.pattr_hdr = hdr;
545 
546 	m->m_data -= hlen;
547 	m->m_len += hlen;
548 
549 	return;
550 
551  bad:
552 	m->m_pkthdr.pattr_class = NULL;
553 	m->m_pkthdr.pattr_hdr = NULL;
554 	m->m_pkthdr.pattr_af = AF_UNSPEC;
555 }
556 #endif /* ALTQ */
557 
558 /*
559  * Process a received Ethernet packet;
560  * the packet is in the mbuf chain m with
561  * the ether header.
562  */
563 void
564 ether_input(struct ifnet *ifp, struct mbuf *m)
565 {
566 	struct ethercom *ec = (struct ethercom *) ifp;
567 	pktqueue_t *pktq = NULL;
568 	struct ifqueue *inq = NULL;
569 	uint16_t etype;
570 	struct ether_header *eh;
571 	size_t ehlen;
572 	static int earlypkts;
573 	int isr = 0;
574 #if defined (LLC) || defined(NETATALK)
575 	struct llc *l;
576 #endif
577 
578 	KASSERT(!cpu_intr_p());
579 
580 	if ((ifp->if_flags & IFF_UP) == 0) {
581 		m_freem(m);
582 		return;
583 	}
584 
585 #ifdef MBUFTRACE
586 	m_claimm(m, &ec->ec_rx_mowner);
587 #endif
588 	eh = mtod(m, struct ether_header *);
589 	etype = ntohs(eh->ether_type);
590 	ehlen = sizeof(*eh);
591 
592 	if(__predict_false(earlypkts < 100 || !rnd_initial_entropy)) {
593 		rnd_add_data(NULL, eh, ehlen, 0);
594 		earlypkts++;
595 	}
596 
597 	/*
598 	 * Determine if the packet is within its size limits.
599 	 */
600 	if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len >
601 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
602 		mutex_enter(&bigpktpps_lock);
603 		if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count,
604 			    bigpktppslim)) {
605 			printf("%s: discarding oversize frame (len=%d)\n",
606 			    ifp->if_xname, m->m_pkthdr.len);
607 		}
608 		mutex_exit(&bigpktpps_lock);
609 		m_freem(m);
610 		return;
611 	}
612 
613 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
614 		/*
615 		 * If this is not a simplex interface, drop the packet
616 		 * if it came from us.
617 		 */
618 		if ((ifp->if_flags & IFF_SIMPLEX) == 0 &&
619 		    memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost,
620 		    ETHER_ADDR_LEN) == 0) {
621 			m_freem(m);
622 			return;
623 		}
624 
625 		if (memcmp(etherbroadcastaddr,
626 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
627 			m->m_flags |= M_BCAST;
628 		else
629 			m->m_flags |= M_MCAST;
630 		ifp->if_imcasts++;
631 	}
632 
633 	/* If the CRC is still on the packet, trim it off. */
634 	if (m->m_flags & M_HASFCS) {
635 		m_adj(m, -ETHER_CRC_LEN);
636 		m->m_flags &= ~M_HASFCS;
637 	}
638 
639 	ifp->if_ibytes += m->m_pkthdr.len;
640 
641 #if NCARP > 0
642 	if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) {
643 		/*
644 		 * clear M_PROMISC, in case the packets comes from a
645 		 * vlan
646 		 */
647 		m->m_flags &= ~M_PROMISC;
648 		if (carp_input(m, (uint8_t *)&eh->ether_shost,
649 		    (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0)
650 			return;
651 	}
652 #endif /* NCARP > 0 */
653 	if ((m->m_flags & (M_BCAST|M_MCAST|M_PROMISC)) == 0 &&
654 	    (ifp->if_flags & IFF_PROMISC) != 0 &&
655 	    memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost,
656 		   ETHER_ADDR_LEN) != 0) {
657 		m->m_flags |= M_PROMISC;
658 	}
659 
660 	if ((m->m_flags & M_PROMISC) == 0) {
661 		if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
662 			return;
663 		if (m == NULL)
664 			return;
665 
666 		eh = mtod(m, struct ether_header *);
667 		etype = ntohs(eh->ether_type);
668 		ehlen = sizeof(*eh);
669 	}
670 
671 #if NAGR > 0
672 	if (ifp->if_agrprivate &&
673 	    __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) {
674 		m->m_flags &= ~M_PROMISC;
675 		agr_input(ifp, m);
676 		return;
677 	}
678 #endif /* NAGR > 0 */
679 
680 	/*
681 	 * If VLANs are configured on the interface, check to
682 	 * see if the device performed the decapsulation and
683 	 * provided us with the tag.
684 	 */
685 	if (ec->ec_nvlans && m_tag_find(m, PACKET_TAG_VLAN, NULL) != NULL) {
686 #if NVLAN > 0
687 		/*
688 		 * vlan_input() will either recursively call ether_input()
689 		 * or drop the packet.
690 		 */
691 		vlan_input(ifp, m);
692 #else
693 		m_freem(m);
694 #endif
695 		return;
696 	}
697 
698 	/*
699 	 * Handle protocols that expect to have the Ethernet header
700 	 * (and possibly FCS) intact.
701 	 */
702 	switch (etype) {
703 	case ETHERTYPE_VLAN: {
704 		struct ether_vlan_header *evl = (void *)eh;
705 		/*
706 		 * If there is a tag of 0, then the VLAN header was probably
707 		 * just being used to store the priority.  Extract the ether
708 		 * type, and if IP or IPV6, let them deal with it.
709 		 */
710 		if (m->m_len <= sizeof(*evl)
711 		    && EVL_VLANOFTAG(evl->evl_tag) == 0) {
712 			etype = ntohs(evl->evl_proto);
713 			ehlen = sizeof(*evl);
714 			if ((m->m_flags & M_PROMISC) == 0
715 			    && (etype == ETHERTYPE_IP
716 				|| etype == ETHERTYPE_IPV6))
717 				break;
718 		}
719 #if NVLAN > 0
720 		/*
721 		 * vlan_input() will either recursively call ether_input()
722 		 * or drop the packet.
723 		 */
724 		if (((struct ethercom *)ifp)->ec_nvlans != 0)
725 			vlan_input(ifp, m);
726 		else
727 #endif /* NVLAN > 0 */
728 			m_freem(m);
729 		return;
730 	}
731 #if NPPPOE > 0
732 	case ETHERTYPE_PPPOEDISC:
733 		pppoedisc_input(ifp, m);
734 		return;
735 	case ETHERTYPE_PPPOE:
736 		pppoe_input(ifp, m);
737 		return;
738 #endif /* NPPPOE > 0 */
739 	case ETHERTYPE_SLOWPROTOCOLS: {
740 		uint8_t subtype;
741 
742 #if defined(DIAGNOSTIC)
743 		if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype)) {
744 			panic("ether_input: too short slow protocol packet");
745 		}
746 #endif
747 		m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype);
748 		switch (subtype) {
749 #if NAGR > 0
750 		case SLOWPROTOCOLS_SUBTYPE_LACP:
751 			if (ifp->if_agrprivate) {
752 				ieee8023ad_lacp_input(ifp, m);
753 				return;
754 			}
755 			break;
756 
757 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
758 			if (ifp->if_agrprivate) {
759 				ieee8023ad_marker_input(ifp, m);
760 				return;
761 			}
762 			break;
763 #endif /* NAGR > 0 */
764 		default:
765 			if (subtype == 0 || subtype > 10) {
766 				/* illegal value */
767 				m_freem(m);
768 				return;
769 			}
770 			/* unknown subtype */
771 			break;
772 		}
773 		/* FALLTHROUGH */
774 	}
775 	default:
776 		if (m->m_flags & M_PROMISC) {
777 			m_freem(m);
778 			return;
779 		}
780 	}
781 
782 	/* If the CRC is still on the packet, trim it off. */
783 	if (m->m_flags & M_HASFCS) {
784 		m_adj(m, -ETHER_CRC_LEN);
785 		m->m_flags &= ~M_HASFCS;
786 	}
787 
788 	if (etype > ETHERMTU + sizeof (struct ether_header)) {
789 		/* Strip off the Ethernet header. */
790 		m_adj(m, ehlen);
791 
792 		switch (etype) {
793 #ifdef INET
794 		case ETHERTYPE_IP:
795 #ifdef GATEWAY
796 			if (ipflow_fastforward(m))
797 				return;
798 #endif
799 			pktq = ip_pktq;
800 			break;
801 
802 		case ETHERTYPE_ARP:
803 			isr = NETISR_ARP;
804 			inq = &arpintrq;
805 			break;
806 
807 		case ETHERTYPE_REVARP:
808 			revarpinput(m);	/* XXX queue? */
809 			return;
810 #endif
811 #ifdef INET6
812 		case ETHERTYPE_IPV6:
813 			if (__predict_false(!in6_present)) {
814 				m_freem(m);
815 				return;
816 			}
817 #ifdef GATEWAY
818 			if (ip6flow_fastforward(&m))
819 				return;
820 #endif
821 			pktq = ip6_pktq;
822 			break;
823 #endif
824 #ifdef NETATALK
825 		case ETHERTYPE_ATALK:
826 			isr = NETISR_ATALK;
827 			inq = &atintrq1;
828 			break;
829 		case ETHERTYPE_AARP:
830 			/* probably this should be done with a NETISR as well */
831 			aarpinput(ifp, m); /* XXX */
832 			return;
833 #endif /* NETATALK */
834 #ifdef MPLS
835 		case ETHERTYPE_MPLS:
836 			isr = NETISR_MPLS;
837 			inq = &mplsintrq;
838 			break;
839 #endif
840 		default:
841 			m_freem(m);
842 			return;
843 		}
844 	} else {
845 #if defined (LLC) || defined (NETATALK)
846 		l = (struct llc *)(eh+1);
847 		switch (l->llc_dsap) {
848 #ifdef NETATALK
849 		case LLC_SNAP_LSAP:
850 			switch (l->llc_control) {
851 			case LLC_UI:
852 				if (l->llc_ssap != LLC_SNAP_LSAP) {
853 					goto dropanyway;
854 				}
855 
856 				if (memcmp(&(l->llc_snap_org_code)[0],
857 				    at_org_code, sizeof(at_org_code)) == 0 &&
858 				    ntohs(l->llc_snap_ether_type) ==
859 				    ETHERTYPE_ATALK) {
860 					inq = &atintrq2;
861 					m_adj(m, sizeof(struct ether_header)
862 					    + sizeof(struct llc));
863 					isr = NETISR_ATALK;
864 					break;
865 				}
866 
867 				if (memcmp(&(l->llc_snap_org_code)[0],
868 				    aarp_org_code,
869 				    sizeof(aarp_org_code)) == 0 &&
870 				    ntohs(l->llc_snap_ether_type) ==
871 				    ETHERTYPE_AARP) {
872 					m_adj( m, sizeof(struct ether_header)
873 					    + sizeof(struct llc));
874 					aarpinput(ifp, m); /* XXX */
875 				    return;
876 				}
877 
878 			default:
879 				goto dropanyway;
880 			}
881 			break;
882 		dropanyway:
883 #endif
884 		default:
885 			m_freem(m);
886 			return;
887 		}
888 #else /* ISO || LLC || NETATALK*/
889 		m_freem(m);
890 		return;
891 #endif /* ISO || LLC || NETATALK*/
892 	}
893 
894 	if (__predict_true(pktq)) {
895 #ifdef NET_MPSAFE
896 		const u_int h = curcpu()->ci_index;
897 #else
898 		const uint32_t h = pktq_rps_hash(m);
899 #endif
900 		if (__predict_false(!pktq_enqueue(pktq, m, h))) {
901 			m_freem(m);
902 		}
903 		return;
904 	}
905 
906 	if (__predict_false(!inq)) {
907 		/* Should not happen. */
908 		m_freem(m);
909 		return;
910 	}
911 
912 	IFQ_LOCK(inq);
913 	if (IF_QFULL(inq)) {
914 		IF_DROP(inq);
915 		IFQ_UNLOCK(inq);
916 		m_freem(m);
917 	} else {
918 		IF_ENQUEUE(inq, m);
919 		IFQ_UNLOCK(inq);
920 		schednetisr(isr);
921 	}
922 }
923 
924 /*
925  * Convert Ethernet address to printable (loggable) representation.
926  */
927 char *
928 ether_sprintf(const u_char *ap)
929 {
930 	static char etherbuf[3 * ETHER_ADDR_LEN];
931 	return ether_snprintf(etherbuf, sizeof(etherbuf), ap);
932 }
933 
934 char *
935 ether_snprintf(char *buf, size_t len, const u_char *ap)
936 {
937 	char *cp = buf;
938 	size_t i;
939 
940 	for (i = 0; i < len / 3; i++) {
941 		*cp++ = hexdigits[*ap >> 4];
942 		*cp++ = hexdigits[*ap++ & 0xf];
943 		*cp++ = ':';
944 	}
945 	*--cp = '\0';
946 	return buf;
947 }
948 
949 /*
950  * Perform common duties while attaching to interface list
951  */
952 void
953 ether_ifattach(struct ifnet *ifp, const uint8_t *lla)
954 {
955 	struct ethercom *ec = (struct ethercom *)ifp;
956 
957 	ifp->if_extflags |= IFEF_OUTPUT_MPSAFE;
958 	ifp->if_type = IFT_ETHER;
959 	ifp->if_hdrlen = ETHER_HDR_LEN;
960 	ifp->if_dlt = DLT_EN10MB;
961 	ifp->if_mtu = ETHERMTU;
962 	ifp->if_output = ether_output;
963 	ifp->_if_input = ether_input;
964 	if (ifp->if_baudrate == 0)
965 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
966 
967 	if (lla != NULL)
968 		if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla));
969 
970 	LIST_INIT(&ec->ec_multiaddrs);
971 	ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
972 	ifp->if_broadcastaddr = etherbroadcastaddr;
973 	bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header));
974 #ifdef MBUFTRACE
975 	strlcpy(ec->ec_tx_mowner.mo_name, ifp->if_xname,
976 	    sizeof(ec->ec_tx_mowner.mo_name));
977 	strlcpy(ec->ec_tx_mowner.mo_descr, "tx",
978 	    sizeof(ec->ec_tx_mowner.mo_descr));
979 	strlcpy(ec->ec_rx_mowner.mo_name, ifp->if_xname,
980 	    sizeof(ec->ec_rx_mowner.mo_name));
981 	strlcpy(ec->ec_rx_mowner.mo_descr, "rx",
982 	    sizeof(ec->ec_rx_mowner.mo_descr));
983 	MOWNER_ATTACH(&ec->ec_tx_mowner);
984 	MOWNER_ATTACH(&ec->ec_rx_mowner);
985 	ifp->if_mowner = &ec->ec_tx_mowner;
986 #endif
987 }
988 
989 void
990 ether_ifdetach(struct ifnet *ifp)
991 {
992 	struct ethercom *ec = (void *) ifp;
993 	struct ether_multi *enm;
994 	int s;
995 
996 	/*
997 	 * Prevent further calls to ioctl (for example turning off
998 	 * promiscuous mode from the bridge code), which eventually can
999 	 * call if_init() which can cause panics because the interface
1000 	 * is in the process of being detached. Return device not configured
1001 	 * instead.
1002 	 */
1003 	ifp->if_ioctl = (int (*)(struct ifnet *, u_long, void *))enxio;
1004 
1005 #if NBRIDGE > 0
1006 	if (ifp->if_bridge)
1007 		bridge_ifdetach(ifp);
1008 #endif
1009 
1010 	bpf_detach(ifp);
1011 
1012 #if NVLAN > 0
1013 	if (ec->ec_nvlans)
1014 		vlan_ifdetach(ifp);
1015 #endif
1016 
1017 	s = splnet();
1018 	mutex_enter(ec->ec_lock);
1019 	while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) {
1020 		LIST_REMOVE(enm, enm_list);
1021 		kmem_free(enm, sizeof(*enm));
1022 		ec->ec_multicnt--;
1023 	}
1024 	mutex_exit(ec->ec_lock);
1025 	splx(s);
1026 
1027 	mutex_destroy(ec->ec_lock);
1028 
1029 	ifp->if_mowner = NULL;
1030 	MOWNER_DETACH(&ec->ec_rx_mowner);
1031 	MOWNER_DETACH(&ec->ec_tx_mowner);
1032 }
1033 
1034 #if 0
1035 /*
1036  * This is for reference.  We have a table-driven version
1037  * of the little-endian crc32 generator, which is faster
1038  * than the double-loop.
1039  */
1040 uint32_t
1041 ether_crc32_le(const uint8_t *buf, size_t len)
1042 {
1043 	uint32_t c, crc, carry;
1044 	size_t i, j;
1045 
1046 	crc = 0xffffffffU;	/* initial value */
1047 
1048 	for (i = 0; i < len; i++) {
1049 		c = buf[i];
1050 		for (j = 0; j < 8; j++) {
1051 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1052 			crc >>= 1;
1053 			c >>= 1;
1054 			if (carry)
1055 				crc = (crc ^ ETHER_CRC_POLY_LE);
1056 		}
1057 	}
1058 
1059 	return (crc);
1060 }
1061 #else
1062 uint32_t
1063 ether_crc32_le(const uint8_t *buf, size_t len)
1064 {
1065 	static const uint32_t crctab[] = {
1066 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1067 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1068 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1069 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1070 	};
1071 	uint32_t crc;
1072 	size_t i;
1073 
1074 	crc = 0xffffffffU;	/* initial value */
1075 
1076 	for (i = 0; i < len; i++) {
1077 		crc ^= buf[i];
1078 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1079 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1080 	}
1081 
1082 	return (crc);
1083 }
1084 #endif
1085 
1086 uint32_t
1087 ether_crc32_be(const uint8_t *buf, size_t len)
1088 {
1089 	uint32_t c, crc, carry;
1090 	size_t i, j;
1091 
1092 	crc = 0xffffffffU;	/* initial value */
1093 
1094 	for (i = 0; i < len; i++) {
1095 		c = buf[i];
1096 		for (j = 0; j < 8; j++) {
1097 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1098 			crc <<= 1;
1099 			c >>= 1;
1100 			if (carry)
1101 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1102 		}
1103 	}
1104 
1105 	return (crc);
1106 }
1107 
1108 #ifdef INET
1109 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] =
1110     { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 };
1111 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] =
1112     { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff };
1113 #endif
1114 #ifdef INET6
1115 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] =
1116     { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 };
1117 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] =
1118     { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff };
1119 #endif
1120 
1121 /*
1122  * ether_aton implementation, not using a static buffer.
1123  */
1124 int
1125 ether_aton_r(u_char *dest, size_t len, const char *str)
1126 {
1127         const u_char *cp = (const void *)str;
1128 	u_char *ep;
1129 
1130 #define atox(c)	(((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10))
1131 
1132 	if (len < ETHER_ADDR_LEN)
1133 		return ENOSPC;
1134 
1135 	ep = dest + ETHER_ADDR_LEN;
1136 
1137 	while (*cp) {
1138                 if (!isxdigit(*cp))
1139                         return EINVAL;
1140 		*dest = atox(*cp);
1141 		cp++;
1142                 if (isxdigit(*cp)) {
1143                         *dest = (*dest << 4) | atox(*cp);
1144 			dest++;
1145 			cp++;
1146                 } else
1147 			dest++;
1148 		if (dest == ep)
1149 			return *cp == '\0' ? 0 : ENAMETOOLONG;
1150 		switch (*cp) {
1151 		case ':':
1152 		case '-':
1153 		case '.':
1154 			cp++;
1155 			break;
1156 		}
1157         }
1158 	return ENOBUFS;
1159 }
1160 
1161 /*
1162  * Convert a sockaddr into an Ethernet address or range of Ethernet
1163  * addresses.
1164  */
1165 int
1166 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN],
1167     uint8_t addrhi[ETHER_ADDR_LEN])
1168 {
1169 #ifdef INET
1170 	const struct sockaddr_in *sin;
1171 #endif /* INET */
1172 #ifdef INET6
1173 	const struct sockaddr_in6 *sin6;
1174 #endif /* INET6 */
1175 
1176 	switch (sa->sa_family) {
1177 
1178 	case AF_UNSPEC:
1179 		memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN);
1180 		memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1181 		break;
1182 
1183 #ifdef INET
1184 	case AF_INET:
1185 		sin = satocsin(sa);
1186 		if (sin->sin_addr.s_addr == INADDR_ANY) {
1187 			/*
1188 			 * An IP address of INADDR_ANY means listen to
1189 			 * or stop listening to all of the Ethernet
1190 			 * multicast addresses used for IP.
1191 			 * (This is for the sake of IP multicast routers.)
1192 			 */
1193 			memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN);
1194 			memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN);
1195 		}
1196 		else {
1197 			ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo);
1198 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1199 		}
1200 		break;
1201 #endif
1202 #ifdef INET6
1203 	case AF_INET6:
1204 		sin6 = satocsin6(sa);
1205 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1206 			/*
1207 			 * An IP6 address of 0 means listen to or stop
1208 			 * listening to all of the Ethernet multicast
1209 			 * address used for IP6.
1210 			 * (This is used for multicast routers.)
1211 			 */
1212 			memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN);
1213 			memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN);
1214 		} else {
1215 			ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo);
1216 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1217 		}
1218 		break;
1219 #endif
1220 
1221 	default:
1222 		return EAFNOSUPPORT;
1223 	}
1224 	return 0;
1225 }
1226 
1227 /*
1228  * Add an Ethernet multicast address or range of addresses to the list for a
1229  * given interface.
1230  */
1231 int
1232 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec)
1233 {
1234 	struct ether_multi *enm, *_enm;
1235 	u_char addrlo[ETHER_ADDR_LEN];
1236 	u_char addrhi[ETHER_ADDR_LEN];
1237 	int s, error = 0;
1238 
1239 	/* Allocate out of lock */
1240 	/* XXX still can be called in softint */
1241 	enm = kmem_intr_alloc(sizeof(*enm), KM_SLEEP);
1242 	if (enm == NULL)
1243 		return ENOBUFS;
1244 
1245 	s = splnet();
1246 	mutex_enter(ec->ec_lock);
1247 	error = ether_multiaddr(sa, addrlo, addrhi);
1248 	if (error != 0)
1249 		goto out;
1250 
1251 	/*
1252 	 * Verify that we have valid Ethernet multicast addresses.
1253 	 */
1254 	if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) {
1255 		error = EINVAL;
1256 		goto out;
1257 	}
1258 	/*
1259 	 * See if the address range is already in the list.
1260 	 */
1261 	ETHER_LOOKUP_MULTI(addrlo, addrhi, ec, _enm);
1262 	if (_enm != NULL) {
1263 		/*
1264 		 * Found it; just increment the reference count.
1265 		 */
1266 		++_enm->enm_refcount;
1267 		error = 0;
1268 		goto out;
1269 	}
1270 	/*
1271 	 * New address or range; malloc a new multicast record
1272 	 * and link it into the interface's multicast list.
1273 	 */
1274 	memcpy(enm->enm_addrlo, addrlo, 6);
1275 	memcpy(enm->enm_addrhi, addrhi, 6);
1276 	enm->enm_refcount = 1;
1277 	LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list);
1278 	ec->ec_multicnt++;
1279 	/*
1280 	 * Return ENETRESET to inform the driver that the list has changed
1281 	 * and its reception filter should be adjusted accordingly.
1282 	 */
1283 	error = ENETRESET;
1284 	enm = NULL;
1285 out:
1286 	mutex_exit(ec->ec_lock);
1287 	splx(s);
1288 	if (enm != NULL)
1289 		kmem_free(enm, sizeof(*enm));
1290 	return error;
1291 }
1292 
1293 /*
1294  * Delete a multicast address record.
1295  */
1296 int
1297 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec)
1298 {
1299 	struct ether_multi *enm;
1300 	u_char addrlo[ETHER_ADDR_LEN];
1301 	u_char addrhi[ETHER_ADDR_LEN];
1302 	int s, error;
1303 
1304 	s = splnet();
1305 	mutex_enter(ec->ec_lock);
1306 	error = ether_multiaddr(sa, addrlo, addrhi);
1307 	if (error != 0)
1308 		goto error;
1309 
1310 	/*
1311 	 * Look ur the address in our list.
1312 	 */
1313 	ETHER_LOOKUP_MULTI(addrlo, addrhi, ec, enm);
1314 	if (enm == NULL) {
1315 		error = ENXIO;
1316 		goto error;
1317 	}
1318 	if (--enm->enm_refcount != 0) {
1319 		/*
1320 		 * Still some claims to this record.
1321 		 */
1322 		error = 0;
1323 		goto error;
1324 	}
1325 	/*
1326 	 * No remaining claims to this record; unlink and free it.
1327 	 */
1328 	LIST_REMOVE(enm, enm_list);
1329 	ec->ec_multicnt--;
1330 	mutex_exit(ec->ec_lock);
1331 	splx(s);
1332 
1333 	kmem_free(enm, sizeof(*enm));
1334 	/*
1335 	 * Return ENETRESET to inform the driver that the list has changed
1336 	 * and its reception filter should be adjusted accordingly.
1337 	 */
1338 	return ENETRESET;
1339 error:
1340 	mutex_exit(ec->ec_lock);
1341 	splx(s);
1342 	return error;
1343 }
1344 
1345 void
1346 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb)
1347 {
1348 	ec->ec_ifflags_cb = cb;
1349 }
1350 
1351 /*
1352  * Common ioctls for Ethernet interfaces.  Note, we must be
1353  * called at splnet().
1354  */
1355 int
1356 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data)
1357 {
1358 	struct ethercom *ec = (void *) ifp;
1359 	struct eccapreq *eccr;
1360 	struct ifreq *ifr = (struct ifreq *)data;
1361 	struct if_laddrreq *iflr = data;
1362 	const struct sockaddr_dl *sdl;
1363 	static const uint8_t zero[ETHER_ADDR_LEN];
1364 	int error;
1365 
1366 	switch (cmd) {
1367 	case SIOCINITIFADDR:
1368 	    {
1369 		struct ifaddr *ifa = (struct ifaddr *)data;
1370 		if (ifa->ifa_addr->sa_family != AF_LINK
1371 		    && (ifp->if_flags & (IFF_UP|IFF_RUNNING)) !=
1372 		       (IFF_UP|IFF_RUNNING)) {
1373 			ifp->if_flags |= IFF_UP;
1374 			if ((error = (*ifp->if_init)(ifp)) != 0)
1375 				return error;
1376 		}
1377 #ifdef INET
1378 		if (ifa->ifa_addr->sa_family == AF_INET)
1379 			arp_ifinit(ifp, ifa);
1380 #endif /* INET */
1381 		return 0;
1382 	    }
1383 
1384 	case SIOCSIFMTU:
1385 	    {
1386 		int maxmtu;
1387 
1388 		if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU)
1389 			maxmtu = ETHERMTU_JUMBO;
1390 		else
1391 			maxmtu = ETHERMTU;
1392 
1393 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu)
1394 			return EINVAL;
1395 		else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET)
1396 			return error;
1397 		else if (ifp->if_flags & IFF_UP) {
1398 			/* Make sure the device notices the MTU change. */
1399 			return (*ifp->if_init)(ifp);
1400 		} else
1401 			return 0;
1402 	    }
1403 
1404 	case SIOCSIFFLAGS:
1405 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
1406 			return error;
1407 		switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
1408 		case IFF_RUNNING:
1409 			/*
1410 			 * If interface is marked down and it is running,
1411 			 * then stop and disable it.
1412 			 */
1413 			(*ifp->if_stop)(ifp, 1);
1414 			break;
1415 		case IFF_UP:
1416 			/*
1417 			 * If interface is marked up and it is stopped, then
1418 			 * start it.
1419 			 */
1420 			return (*ifp->if_init)(ifp);
1421 		case IFF_UP|IFF_RUNNING:
1422 			error = 0;
1423 			if (ec->ec_ifflags_cb == NULL ||
1424 			    (error = (*ec->ec_ifflags_cb)(ec)) == ENETRESET) {
1425 				/*
1426 				 * Reset the interface to pick up
1427 				 * changes in any other flags that
1428 				 * affect the hardware state.
1429 				 */
1430 				return (*ifp->if_init)(ifp);
1431 			} else
1432 				return error;
1433 		case 0:
1434 			break;
1435 		}
1436 		return 0;
1437 	case SIOCGETHERCAP:
1438 		eccr = (struct eccapreq *)data;
1439 		eccr->eccr_capabilities = ec->ec_capabilities;
1440 		eccr->eccr_capenable = ec->ec_capenable;
1441 		return 0;
1442 	case SIOCADDMULTI:
1443 		return ether_addmulti(ifreq_getaddr(cmd, ifr), ec);
1444 	case SIOCDELMULTI:
1445 		return ether_delmulti(ifreq_getaddr(cmd, ifr), ec);
1446 	case SIOCSIFMEDIA:
1447 	case SIOCGIFMEDIA:
1448 		if (ec->ec_mii == NULL)
1449 			return ENOTTY;
1450 		return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media, cmd);
1451 	case SIOCALIFADDR:
1452 		sdl = satocsdl(sstocsa(&iflr->addr));
1453 		if (sdl->sdl_family != AF_LINK)
1454 			;
1455 		else if (ETHER_IS_MULTICAST(CLLADDR(sdl)))
1456 			return EINVAL;
1457 		else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0)
1458 			return EINVAL;
1459 		/*FALLTHROUGH*/
1460 	default:
1461 		return ifioctl_common(ifp, cmd, data);
1462 	}
1463 	return 0;
1464 }
1465 
1466 /*
1467  * Enable/disable passing VLAN packets if the parent interface supports it.
1468  * Return:
1469  * 	 0: Ok
1470  *	-1: Parent interface does not support vlans
1471  *	>0: Error
1472  */
1473 int
1474 ether_enable_vlan_mtu(struct ifnet *ifp)
1475 {
1476 	int error;
1477 	struct ethercom *ec = (void *)ifp;
1478 
1479 	/* Parent does not support VLAN's */
1480 	if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0)
1481 		return -1;
1482 
1483 	/*
1484 	 * Parent supports the VLAN_MTU capability,
1485 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames;
1486 	 * enable it.
1487 	 */
1488 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1489 
1490 	/* Interface is down, defer for later */
1491 	if ((ifp->if_flags & IFF_UP) == 0)
1492 		return 0;
1493 
1494 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1495 		return 0;
1496 
1497 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1498 	return error;
1499 }
1500 
1501 int
1502 ether_disable_vlan_mtu(struct ifnet *ifp)
1503 {
1504 	int error;
1505 	struct ethercom *ec = (void *)ifp;
1506 
1507 	/* We still have VLAN's, defer for later */
1508 	if (ec->ec_nvlans != 0)
1509 		return 0;
1510 
1511 	/* Parent does not support VLAB's, nothing to do. */
1512 	if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0)
1513 		return -1;
1514 
1515 	/*
1516 	 * Disable Tx/Rx of VLAN-sized frames.
1517 	 */
1518 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1519 
1520 	/* Interface is down, defer for later */
1521 	if ((ifp->if_flags & IFF_UP) == 0)
1522 		return 0;
1523 
1524 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1525 		return 0;
1526 
1527 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1528 	return error;
1529 }
1530 
1531 static int
1532 ether_multicast_sysctl(SYSCTLFN_ARGS)
1533 {
1534 	struct ether_multi *enm;
1535 	struct ifnet *ifp;
1536 	struct ethercom *ec;
1537 	int error = 0;
1538 	size_t written;
1539 	struct psref psref;
1540 	int bound, s;
1541 	unsigned int multicnt;
1542 	struct ether_multi_sysctl *addrs;
1543 	int i;
1544 
1545 	if (namelen != 1)
1546 		return EINVAL;
1547 
1548 	bound = curlwp_bind();
1549 	ifp = if_get_byindex(name[0], &psref);
1550 	if (ifp == NULL) {
1551 		error = ENODEV;
1552 		goto out;
1553 	}
1554 	if (ifp->if_type != IFT_ETHER) {
1555 		if_put(ifp, &psref);
1556 		*oldlenp = 0;
1557 		goto out;
1558 	}
1559 	ec = (struct ethercom *)ifp;
1560 
1561 	if (oldp == NULL) {
1562 		if_put(ifp, &psref);
1563 		*oldlenp = ec->ec_multicnt * sizeof(*addrs);
1564 		goto out;
1565 	}
1566 
1567 	/*
1568 	 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which
1569 	 * is sleepable, with holding it. Copy data to a local buffer first
1570 	 * with holding it and then call sysctl_copyout without holding it.
1571 	 */
1572 retry:
1573 	multicnt = ec->ec_multicnt;
1574 	addrs = kmem_alloc(sizeof(*addrs) * multicnt, KM_SLEEP);
1575 
1576 	s = splnet();
1577 	mutex_enter(ec->ec_lock);
1578 	if (multicnt < ec->ec_multicnt) {
1579 		/* The number of multicast addresses have increased */
1580 		mutex_exit(ec->ec_lock);
1581 		splx(s);
1582 		kmem_free(addrs, sizeof(*addrs) * multicnt);
1583 		goto retry;
1584 	}
1585 
1586 	i = 0;
1587 	LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) {
1588 		struct ether_multi_sysctl *addr = &addrs[i];
1589 		addr->enm_refcount = enm->enm_refcount;
1590 		memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN);
1591 		memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN);
1592 		i++;
1593 	}
1594 	mutex_exit(ec->ec_lock);
1595 	splx(s);
1596 
1597 	error = 0;
1598 	written = 0;
1599 	for (i = 0; i < multicnt; i++) {
1600 		struct ether_multi_sysctl *addr = &addrs[i];
1601 
1602 		if (written + sizeof(*addr) > *oldlenp)
1603 			break;
1604 		error = sysctl_copyout(l, addr, oldp, sizeof(*addr));
1605 		if (error)
1606 			break;
1607 		written += sizeof(*addr);
1608 		oldp = (char *)oldp + sizeof(*addr);
1609 	}
1610 	kmem_free(addrs, sizeof(*addrs) * multicnt);
1611 
1612 	if_put(ifp, &psref);
1613 
1614 	*oldlenp = written;
1615 out:
1616 	curlwp_bindx(bound);
1617 	return error;
1618 }
1619 
1620 static void
1621 ether_sysctl_setup(struct sysctllog **clog)
1622 {
1623 	const struct sysctlnode *rnode = NULL;
1624 
1625 	sysctl_createv(clog, 0, NULL, &rnode,
1626 		       CTLFLAG_PERMANENT,
1627 		       CTLTYPE_NODE, "ether",
1628 		       SYSCTL_DESCR("Ethernet-specific information"),
1629 		       NULL, 0, NULL, 0,
1630 		       CTL_NET, CTL_CREATE, CTL_EOL);
1631 
1632 	sysctl_createv(clog, 0, &rnode, NULL,
1633 		       CTLFLAG_PERMANENT,
1634 		       CTLTYPE_NODE, "multicast",
1635 		       SYSCTL_DESCR("multicast addresses"),
1636 		       ether_multicast_sysctl, 0, NULL, 0,
1637 		       CTL_CREATE, CTL_EOL);
1638 }
1639 
1640 void
1641 etherinit(void)
1642 {
1643 
1644 	mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET);
1645 	ether_sysctl_setup(NULL);
1646 }
1647