xref: /netbsd-src/sys/net/if_ethersubr.c (revision 181254a7b1bdde6873432bffef2d2decc4b5c22f)
1 /*	$NetBSD: if_ethersubr.c,v 1.288 2020/08/28 06:27:49 ozaki-r 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 1.288 2020/08/28 06:27:49 ozaki-r Exp $");
65 
66 #ifdef _KERNEL_OPT
67 #include "opt_inet.h"
68 #include "opt_atalk.h"
69 #include "opt_mbuftrace.h"
70 #include "opt_mpls.h"
71 #include "opt_gateway.h"
72 #include "opt_pppoe.h"
73 #include "opt_net_mpsafe.h"
74 #endif
75 
76 #include "vlan.h"
77 #include "pppoe.h"
78 #include "bridge.h"
79 #include "arp.h"
80 #include "agr.h"
81 
82 #include <sys/sysctl.h>
83 #include <sys/mbuf.h>
84 #include <sys/mutex.h>
85 #include <sys/ioctl.h>
86 #include <sys/errno.h>
87 #include <sys/device.h>
88 #include <sys/entropy.h>
89 #include <sys/rndsource.h>
90 #include <sys/cpu.h>
91 #include <sys/kmem.h>
92 
93 #include <net/if.h>
94 #include <net/netisr.h>
95 #include <net/route.h>
96 #include <net/if_llc.h>
97 #include <net/if_dl.h>
98 #include <net/if_types.h>
99 #include <net/pktqueue.h>
100 
101 #include <net/if_media.h>
102 #include <dev/mii/mii.h>
103 #include <dev/mii/miivar.h>
104 
105 #if NARP == 0
106 /*
107  * XXX there should really be a way to issue this warning from within config(8)
108  */
109 #error You have included NETATALK or a pseudo-device in your configuration that depends on the presence of ethernet interfaces, but have no such interfaces configured. Check if you really need pseudo-device bridge, pppoe, vlan or options NETATALK.
110 #endif
111 
112 #include <net/bpf.h>
113 
114 #include <net/if_ether.h>
115 #include <net/if_vlanvar.h>
116 
117 #if NPPPOE > 0
118 #include <net/if_pppoe.h>
119 #endif
120 
121 #if NAGR > 0
122 #include <net/agr/ieee8023_slowprotocols.h>	/* XXX */
123 #include <net/agr/ieee8023ad.h>
124 #include <net/agr/if_agrvar.h>
125 #endif
126 
127 #if NBRIDGE > 0
128 #include <net/if_bridgevar.h>
129 #endif
130 
131 #include <netinet/in.h>
132 #ifdef INET
133 #include <netinet/in_var.h>
134 #endif
135 #include <netinet/if_inarp.h>
136 
137 #ifdef INET6
138 #ifndef INET
139 #include <netinet/in.h>
140 #endif
141 #include <netinet6/in6_var.h>
142 #include <netinet6/nd6.h>
143 #endif
144 
145 #include "carp.h"
146 #if NCARP > 0
147 #include <netinet/ip_carp.h>
148 #endif
149 
150 #ifdef NETATALK
151 #include <netatalk/at.h>
152 #include <netatalk/at_var.h>
153 #include <netatalk/at_extern.h>
154 
155 #define llc_snap_org_code llc_un.type_snap.org_code
156 #define llc_snap_ether_type llc_un.type_snap.ether_type
157 
158 extern u_char	at_org_code[3];
159 extern u_char	aarp_org_code[3];
160 #endif /* NETATALK */
161 
162 #ifdef MPLS
163 #include <netmpls/mpls.h>
164 #include <netmpls/mpls_var.h>
165 #endif
166 
167 #ifdef DIAGNOSTIC
168 static struct timeval bigpktppslim_last;
169 static int bigpktppslim = 2;	/* XXX */
170 static int bigpktpps_count;
171 static kmutex_t bigpktpps_lock __cacheline_aligned;
172 #endif
173 
174 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] =
175     { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
176 const uint8_t ethermulticastaddr_slowprotocols[ETHER_ADDR_LEN] =
177     { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x02 };
178 #define senderr(e) { error = (e); goto bad;}
179 
180 static int ether_output(struct ifnet *, struct mbuf *,
181     const struct sockaddr *, const struct rtentry *);
182 
183 /*
184  * Ethernet output routine.
185  * Encapsulate a packet of type family for the local net.
186  * Assumes that ifp is actually pointer to ethercom structure.
187  */
188 static int
189 ether_output(struct ifnet * const ifp0, struct mbuf * const m0,
190     const struct sockaddr * const dst, const struct rtentry *rt)
191 {
192 	uint8_t esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
193 	uint16_t etype = 0;
194 	int error = 0, hdrcmplt = 0;
195 	struct mbuf *m = m0;
196 	struct mbuf *mcopy = NULL;
197 	struct ether_header *eh;
198 	struct ifnet *ifp = ifp0;
199 #ifdef INET
200 	struct arphdr *ah;
201 #endif
202 #ifdef NETATALK
203 	struct at_ifaddr *aa;
204 #endif
205 
206 #ifdef MBUFTRACE
207 	m_claimm(m, ifp->if_mowner);
208 #endif
209 
210 #if NCARP > 0
211 	if (ifp->if_type == IFT_CARP) {
212 		struct ifaddr *ifa;
213 		int s = pserialize_read_enter();
214 
215 		/* loop back if this is going to the carp interface */
216 		if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP &&
217 		    (ifa = ifa_ifwithaddr(dst)) != NULL) {
218 			if (ifa->ifa_ifp == ifp0) {
219 				pserialize_read_exit(s);
220 				return looutput(ifp0, m, dst, rt);
221 			}
222 		}
223 		pserialize_read_exit(s);
224 
225 		ifp = ifp->if_carpdev;
226 		/* ac = (struct arpcom *)ifp; */
227 
228 		if ((ifp0->if_flags & (IFF_UP | IFF_RUNNING)) !=
229 		    (IFF_UP | IFF_RUNNING))
230 			senderr(ENETDOWN);
231 	}
232 #endif
233 
234 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
235 		senderr(ENETDOWN);
236 
237 	switch (dst->sa_family) {
238 
239 #ifdef INET
240 	case AF_INET:
241 		if (m->m_flags & M_BCAST) {
242 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
243 		} else if (m->m_flags & M_MCAST) {
244 			ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst);
245 		} else {
246 			error = arpresolve(ifp0, rt, m, dst, edst, sizeof(edst));
247 			if (error)
248 				return (error == EWOULDBLOCK) ? 0 : error;
249 		}
250 		/* If broadcasting on a simplex interface, loopback a copy */
251 		if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
252 			mcopy = m_copypacket(m, M_DONTWAIT);
253 		etype = htons(ETHERTYPE_IP);
254 		break;
255 
256 	case AF_ARP:
257 		ah = mtod(m, struct arphdr *);
258 		if (m->m_flags & M_BCAST) {
259 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
260 		} else {
261 			void *tha = ar_tha(ah);
262 
263 			if (tha == NULL) {
264 				/* fake with ARPHRD_IEEE1394 */
265 				m_freem(m);
266 				return 0;
267 			}
268 			memcpy(edst, tha, sizeof(edst));
269 		}
270 
271 		ah->ar_hrd = htons(ARPHRD_ETHER);
272 
273 		switch (ntohs(ah->ar_op)) {
274 		case ARPOP_REVREQUEST:
275 		case ARPOP_REVREPLY:
276 			etype = htons(ETHERTYPE_REVARP);
277 			break;
278 
279 		case ARPOP_REQUEST:
280 		case ARPOP_REPLY:
281 		default:
282 			etype = htons(ETHERTYPE_ARP);
283 		}
284 		break;
285 #endif
286 
287 #ifdef INET6
288 	case AF_INET6:
289 		if (m->m_flags & M_BCAST) {
290 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
291 		} else if (m->m_flags & M_MCAST) {
292 			ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr,
293 			    edst);
294 		} else {
295 			error = nd6_resolve(ifp0, rt, m, dst, edst,
296 			    sizeof(edst));
297 			if (error)
298 				return (error == EWOULDBLOCK) ? 0 : error;
299 		}
300 		etype = htons(ETHERTYPE_IPV6);
301 		break;
302 #endif
303 
304 #ifdef NETATALK
305 	case AF_APPLETALK: {
306 		struct ifaddr *ifa;
307 		int s;
308 
309 		KERNEL_LOCK(1, NULL);
310 
311 		if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) {
312 			KERNEL_UNLOCK_ONE(NULL);
313 			return 0;
314 		}
315 
316 		/*
317 		 * ifaddr is the first thing in at_ifaddr
318 		 */
319 		s = pserialize_read_enter();
320 		ifa = at_ifawithnet((const struct sockaddr_at *)dst, ifp);
321 		if (ifa == NULL) {
322 			pserialize_read_exit(s);
323 			KERNEL_UNLOCK_ONE(NULL);
324 			senderr(EADDRNOTAVAIL);
325 		}
326 		aa = (struct at_ifaddr *)ifa;
327 
328 		/*
329 		 * In the phase 2 case, we need to prepend an mbuf for the
330 		 * llc header.
331 		 */
332 		if (aa->aa_flags & AFA_PHASE2) {
333 			struct llc llc;
334 
335 			M_PREPEND(m, sizeof(struct llc), M_DONTWAIT);
336 			if (m == NULL) {
337 				pserialize_read_exit(s);
338 				KERNEL_UNLOCK_ONE(NULL);
339 				senderr(ENOBUFS);
340 			}
341 
342 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
343 			llc.llc_control = LLC_UI;
344 			memcpy(llc.llc_snap_org_code, at_org_code,
345 			    sizeof(llc.llc_snap_org_code));
346 			llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK);
347 			memcpy(mtod(m, void *), &llc, sizeof(struct llc));
348 		} else {
349 			etype = htons(ETHERTYPE_ATALK);
350 		}
351 		pserialize_read_exit(s);
352 		KERNEL_UNLOCK_ONE(NULL);
353 		break;
354 	}
355 #endif /* NETATALK */
356 
357 	case pseudo_AF_HDRCMPLT:
358 		hdrcmplt = 1;
359 		memcpy(esrc,
360 		    ((const struct ether_header *)dst->sa_data)->ether_shost,
361 		    sizeof(esrc));
362 		/* FALLTHROUGH */
363 
364 	case AF_UNSPEC:
365 		memcpy(edst,
366 		    ((const struct ether_header *)dst->sa_data)->ether_dhost,
367 		    sizeof(edst));
368 		/* AF_UNSPEC doesn't swap the byte order of the ether_type. */
369 		etype = ((const struct ether_header *)dst->sa_data)->ether_type;
370 		break;
371 
372 	default:
373 		printf("%s: can't handle af%d\n", ifp->if_xname,
374 		    dst->sa_family);
375 		senderr(EAFNOSUPPORT);
376 	}
377 
378 #ifdef MPLS
379 	{
380 		struct m_tag *mtag;
381 		mtag = m_tag_find(m, PACKET_TAG_MPLS);
382 		if (mtag != NULL) {
383 			/* Having the tag itself indicates it's MPLS */
384 			etype = htons(ETHERTYPE_MPLS);
385 			m_tag_delete(m, mtag);
386 		}
387 	}
388 #endif
389 
390 	if (mcopy)
391 		(void)looutput(ifp, mcopy, dst, rt);
392 
393 	KASSERT((m->m_flags & M_PKTHDR) != 0);
394 
395 	/*
396 	 * If no ether type is set, this must be a 802.2 formatted packet.
397 	 */
398 	if (etype == 0)
399 		etype = htons(m->m_pkthdr.len);
400 
401 	/*
402 	 * Add local net header. If no space in first mbuf, allocate another.
403 	 */
404 	M_PREPEND(m, sizeof(struct ether_header), M_DONTWAIT);
405 	if (m == NULL)
406 		senderr(ENOBUFS);
407 
408 	eh = mtod(m, struct ether_header *);
409 	/* Note: etype is already in network byte order. */
410 	memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
411 	memcpy(eh->ether_dhost, edst, sizeof(edst));
412 	if (hdrcmplt) {
413 		memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost));
414 	} else {
415 	 	memcpy(eh->ether_shost, CLLADDR(ifp->if_sadl),
416 		    sizeof(eh->ether_shost));
417 	}
418 
419 #if NCARP > 0
420 	if (ifp0 != ifp && ifp0->if_type == IFT_CARP) {
421 	 	memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl),
422 		    sizeof(eh->ether_shost));
423 	}
424 #endif
425 
426 	if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0)
427 		return error;
428 	if (m == NULL)
429 		return 0;
430 
431 #if NBRIDGE > 0
432 	/*
433 	 * Bridges require special output handling.
434 	 */
435 	if (ifp->if_bridge)
436 		return bridge_output(ifp, m, NULL, NULL);
437 #endif
438 
439 #if NCARP > 0
440 	if (ifp != ifp0)
441 		if_statadd(ifp0, if_obytes, m->m_pkthdr.len + ETHER_HDR_LEN);
442 #endif
443 
444 #ifdef ALTQ
445 	KERNEL_LOCK(1, NULL);
446 	/*
447 	 * If ALTQ is enabled on the parent interface, do
448 	 * classification; the queueing discipline might not
449 	 * require classification, but might require the
450 	 * address family/header pointer in the pktattr.
451 	 */
452 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
453 		altq_etherclassify(&ifp->if_snd, m);
454 	KERNEL_UNLOCK_ONE(NULL);
455 #endif
456 	return ifq_enqueue(ifp, m);
457 
458 bad:
459 	if_statinc(ifp, if_oerrors);
460 	if (m)
461 		m_freem(m);
462 	return error;
463 }
464 
465 #ifdef ALTQ
466 /*
467  * This routine is a slight hack to allow a packet to be classified
468  * if the Ethernet headers are present.  It will go away when ALTQ's
469  * classification engine understands link headers.
470  *
471  * XXX: We may need to do m_pullups here. First to ensure struct ether_header
472  * is indeed contiguous, then to read the LLC and so on.
473  */
474 void
475 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m)
476 {
477 	struct ether_header *eh;
478 	struct mbuf *mtop = m;
479 	uint16_t ether_type;
480 	int hlen, af, hdrsize;
481 	void *hdr;
482 
483 	KASSERT((mtop->m_flags & M_PKTHDR) != 0);
484 
485 	hlen = ETHER_HDR_LEN;
486 	eh = mtod(m, struct ether_header *);
487 
488 	ether_type = htons(eh->ether_type);
489 
490 	if (ether_type < ETHERMTU) {
491 		/* LLC/SNAP */
492 		struct llc *llc = (struct llc *)(eh + 1);
493 		hlen += 8;
494 
495 		if (m->m_len < hlen ||
496 		    llc->llc_dsap != LLC_SNAP_LSAP ||
497 		    llc->llc_ssap != LLC_SNAP_LSAP ||
498 		    llc->llc_control != LLC_UI) {
499 			/* Not SNAP. */
500 			goto bad;
501 		}
502 
503 		ether_type = htons(llc->llc_un.type_snap.ether_type);
504 	}
505 
506 	switch (ether_type) {
507 	case ETHERTYPE_IP:
508 		af = AF_INET;
509 		hdrsize = 20;		/* sizeof(struct ip) */
510 		break;
511 
512 	case ETHERTYPE_IPV6:
513 		af = AF_INET6;
514 		hdrsize = 40;		/* sizeof(struct ip6_hdr) */
515 		break;
516 
517 	default:
518 		af = AF_UNSPEC;
519 		hdrsize = 0;
520 		break;
521 	}
522 
523 	while (m->m_len <= hlen) {
524 		hlen -= m->m_len;
525 		m = m->m_next;
526 		if (m == NULL)
527 			goto bad;
528 	}
529 
530 	if (m->m_len < (hlen + hdrsize)) {
531 		/*
532 		 * protocol header not in a single mbuf.
533 		 * We can't cope with this situation right
534 		 * now (but it shouldn't ever happen, really, anyhow).
535 		 */
536 #ifdef DEBUG
537 		printf("altq_etherclassify: headers span multiple mbufs: "
538 		    "%d < %d\n", m->m_len, (hlen + hdrsize));
539 #endif
540 		goto bad;
541 	}
542 
543 	m->m_data += hlen;
544 	m->m_len -= hlen;
545 
546 	hdr = mtod(m, void *);
547 
548 	if (ALTQ_NEEDS_CLASSIFY(ifq)) {
549 		mtop->m_pkthdr.pattr_class =
550 		    (*ifq->altq_classify)(ifq->altq_clfier, m, af);
551 	}
552 	mtop->m_pkthdr.pattr_af = af;
553 	mtop->m_pkthdr.pattr_hdr = hdr;
554 
555 	m->m_data -= hlen;
556 	m->m_len += hlen;
557 
558 	return;
559 
560 bad:
561 	mtop->m_pkthdr.pattr_class = NULL;
562 	mtop->m_pkthdr.pattr_hdr = NULL;
563 	mtop->m_pkthdr.pattr_af = AF_UNSPEC;
564 }
565 #endif /* ALTQ */
566 
567 #if defined (LLC) || defined (NETATALK)
568 static void
569 ether_input_llc(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh)
570 {
571 	struct ifqueue *inq = NULL;
572 	int isr = 0;
573 	struct llc *l;
574 
575 	if (m->m_len < sizeof(*eh) + sizeof(struct llc))
576 		goto drop;
577 
578 	l = (struct llc *)(eh+1);
579 	switch (l->llc_dsap) {
580 #ifdef NETATALK
581 	case LLC_SNAP_LSAP:
582 		switch (l->llc_control) {
583 		case LLC_UI:
584 			if (l->llc_ssap != LLC_SNAP_LSAP)
585 				goto drop;
586 
587 			if (memcmp(&(l->llc_snap_org_code)[0],
588 			    at_org_code, sizeof(at_org_code)) == 0 &&
589 			    ntohs(l->llc_snap_ether_type) ==
590 			    ETHERTYPE_ATALK) {
591 				inq = &atintrq2;
592 				m_adj(m, sizeof(struct ether_header)
593 				    + sizeof(struct llc));
594 				isr = NETISR_ATALK;
595 				break;
596 			}
597 
598 			if (memcmp(&(l->llc_snap_org_code)[0],
599 			    aarp_org_code,
600 			    sizeof(aarp_org_code)) == 0 &&
601 			    ntohs(l->llc_snap_ether_type) ==
602 			    ETHERTYPE_AARP) {
603 				m_adj(m, sizeof(struct ether_header)
604 				    + sizeof(struct llc));
605 				aarpinput(ifp, m); /* XXX queue? */
606 				return;
607 			}
608 
609 		default:
610 			goto drop;
611 		}
612 		break;
613 #endif
614 	default:
615 		goto drop;
616 	}
617 
618 	KASSERT(inq != NULL);
619 	IFQ_ENQUEUE_ISR(inq, m, isr);
620 	return;
621 
622 drop:
623 	m_freem(m);
624 	if_statinc(ifp, if_ierrors); /* XXX should have a dedicated counter? */
625 	return;
626 }
627 #endif /* defined (LLC) || defined (NETATALK) */
628 
629 /*
630  * Process a received Ethernet packet;
631  * the packet is in the mbuf chain m with
632  * the ether header.
633  */
634 void
635 ether_input(struct ifnet *ifp, struct mbuf *m)
636 {
637 	struct ethercom *ec = (struct ethercom *) ifp;
638 	pktqueue_t *pktq = NULL;
639 	struct ifqueue *inq = NULL;
640 	uint16_t etype;
641 	struct ether_header *eh;
642 	size_t ehlen;
643 	static int earlypkts;
644 	int isr = 0;
645 
646 	KASSERT(!cpu_intr_p());
647 	KASSERT((m->m_flags & M_PKTHDR) != 0);
648 
649 	if ((ifp->if_flags & IFF_UP) == 0)
650 		goto drop;
651 	if (m->m_len < sizeof(*eh)) {
652 		m = m_pullup(m, sizeof(*eh));
653 		if (m == NULL)
654 			goto dropped;
655 	}
656 
657 #ifdef MBUFTRACE
658 	m_claimm(m, &ec->ec_rx_mowner);
659 #endif
660 	eh = mtod(m, struct ether_header *);
661 	etype = ntohs(eh->ether_type);
662 	ehlen = sizeof(*eh);
663 
664 	if (__predict_false(earlypkts < 100 ||
665 		entropy_epoch() == (unsigned)-1)) {
666 		rnd_add_data(NULL, eh, ehlen, 0);
667 		earlypkts++;
668 	}
669 
670 	/*
671 	 * Determine if the packet is within its size limits. For MPLS the
672 	 * header length is variable, so we skip the check.
673 	 */
674 	if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len >
675 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
676 #ifdef DIAGNOSTIC
677 		mutex_enter(&bigpktpps_lock);
678 		if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count,
679 		    bigpktppslim)) {
680 			printf("%s: discarding oversize frame (len=%d)\n",
681 			    ifp->if_xname, m->m_pkthdr.len);
682 		}
683 		mutex_exit(&bigpktpps_lock);
684 #endif
685 		goto drop;
686 	}
687 
688 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
689 		/*
690 		 * If this is not a simplex interface, drop the packet
691 		 * if it came from us.
692 		 */
693 		if ((ifp->if_flags & IFF_SIMPLEX) == 0 &&
694 		    memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost,
695 		    ETHER_ADDR_LEN) == 0) {
696 			goto drop;
697 		}
698 
699 		if (memcmp(etherbroadcastaddr,
700 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
701 			m->m_flags |= M_BCAST;
702 		else
703 			m->m_flags |= M_MCAST;
704 		if_statinc(ifp, if_imcasts);
705 	}
706 
707 	/* If the CRC is still on the packet, trim it off. */
708 	if (m->m_flags & M_HASFCS) {
709 		m_adj(m, -ETHER_CRC_LEN);
710 		m->m_flags &= ~M_HASFCS;
711 	}
712 
713 	if_statadd(ifp, if_ibytes, m->m_pkthdr.len);
714 
715 #if NCARP > 0
716 	if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) {
717 		/*
718 		 * Clear M_PROMISC, in case the packet comes from a
719 		 * vlan.
720 		 */
721 		m->m_flags &= ~M_PROMISC;
722 		if (carp_input(m, (uint8_t *)&eh->ether_shost,
723 		    (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0)
724 			return;
725 	}
726 #endif
727 
728 	if ((m->m_flags & (M_BCAST | M_MCAST | M_PROMISC)) == 0 &&
729 	    (ifp->if_flags & IFF_PROMISC) != 0 &&
730 	    memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost,
731 	     ETHER_ADDR_LEN) != 0) {
732 		m->m_flags |= M_PROMISC;
733 	}
734 
735 	if ((m->m_flags & M_PROMISC) == 0) {
736 		if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
737 			return;
738 		if (m == NULL)
739 			return;
740 
741 		eh = mtod(m, struct ether_header *);
742 		etype = ntohs(eh->ether_type);
743 	}
744 
745 #if NAGR > 0
746 	if (ifp->if_agrprivate &&
747 	    __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) {
748 		m->m_flags &= ~M_PROMISC;
749 		agr_input(ifp, m);
750 		return;
751 	}
752 #endif
753 
754 	/*
755 	 * If VLANs are configured on the interface, check to
756 	 * see if the device performed the decapsulation and
757 	 * provided us with the tag.
758 	 */
759 	if (ec->ec_nvlans && vlan_has_tag(m)) {
760 #if NVLAN > 0
761 		/*
762 		 * vlan_input() will either recursively call ether_input()
763 		 * or drop the packet.
764 		 */
765 		vlan_input(ifp, m);
766 		return;
767 #else
768 		goto drop;
769 #endif
770 	}
771 
772 	/*
773 	 * Handle protocols that expect to have the Ethernet header
774 	 * (and possibly FCS) intact.
775 	 */
776 	switch (etype) {
777 	case ETHERTYPE_VLAN: {
778 		struct ether_vlan_header *evl = (void *)eh;
779 
780 		/*
781 		 * If there is a tag of 0, then the VLAN header was probably
782 		 * just being used to store the priority.  Extract the ether
783 		 * type, and if IP or IPV6, let them deal with it.
784 		 */
785 		if (m->m_len >= sizeof(*evl) &&
786 		    EVL_VLANOFTAG(evl->evl_tag) == 0) {
787 			etype = ntohs(evl->evl_proto);
788 			ehlen = sizeof(*evl);
789 			if ((m->m_flags & M_PROMISC) == 0 &&
790 			    (etype == ETHERTYPE_IP ||
791 			     etype == ETHERTYPE_IPV6))
792 				break;
793 		}
794 
795 #if NVLAN > 0
796 		/*
797 		 * vlan_input() will either recursively call ether_input()
798 		 * or drop the packet.
799 		 */
800 		if (ec->ec_nvlans != 0) {
801 			vlan_input(ifp, m);
802 			return;
803 		} else
804 #endif
805 			goto drop;
806 	}
807 
808 #if NPPPOE > 0
809 	case ETHERTYPE_PPPOEDISC:
810 		pppoedisc_input(ifp, m);
811 		return;
812 
813 	case ETHERTYPE_PPPOE:
814 		pppoe_input(ifp, m);
815 		return;
816 #endif
817 
818 	case ETHERTYPE_SLOWPROTOCOLS: {
819 		uint8_t subtype;
820 
821 		if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype))
822 			goto drop;
823 
824 		m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype);
825 		switch (subtype) {
826 #if NAGR > 0
827 		case SLOWPROTOCOLS_SUBTYPE_LACP:
828 			if (ifp->if_agrprivate) {
829 				ieee8023ad_lacp_input(ifp, m);
830 				return;
831 			}
832 			break;
833 
834 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
835 			if (ifp->if_agrprivate) {
836 				ieee8023ad_marker_input(ifp, m);
837 				return;
838 			}
839 			break;
840 #endif
841 
842 		default:
843 			if (subtype == 0 || subtype > 10) {
844 				/* illegal value */
845 				goto drop;
846 			}
847 			/* unknown subtype */
848 			break;
849 		}
850 	}
851 	/* FALLTHROUGH */
852 	default:
853 		if (m->m_flags & M_PROMISC)
854 			goto drop;
855 	}
856 
857 	/* If the CRC is still on the packet, trim it off. */
858 	if (m->m_flags & M_HASFCS) {
859 		m_adj(m, -ETHER_CRC_LEN);
860 		m->m_flags &= ~M_HASFCS;
861 	}
862 
863 	/* etype represents the size of the payload in this case */
864 	if (etype <= ETHERMTU + sizeof(struct ether_header)) {
865 		KASSERT(ehlen == sizeof(*eh));
866 #if defined (LLC) || defined (NETATALK)
867 		ether_input_llc(ifp, m, eh);
868 		return;
869 #else
870 		goto drop;
871 #endif
872 	}
873 
874 	/* Strip off the Ethernet header. */
875 	m_adj(m, ehlen);
876 
877 	switch (etype) {
878 #ifdef INET
879 	case ETHERTYPE_IP:
880 #ifdef GATEWAY
881 		if (ipflow_fastforward(m))
882 			return;
883 #endif
884 		pktq = ip_pktq;
885 		break;
886 
887 	case ETHERTYPE_ARP:
888 		isr = NETISR_ARP;
889 		inq = &arpintrq;
890 		break;
891 
892 	case ETHERTYPE_REVARP:
893 		revarpinput(m);	/* XXX queue? */
894 		return;
895 #endif
896 
897 #ifdef INET6
898 	case ETHERTYPE_IPV6:
899 		if (__predict_false(!in6_present))
900 			goto drop;
901 #ifdef GATEWAY
902 		if (ip6flow_fastforward(&m))
903 			return;
904 #endif
905 		pktq = ip6_pktq;
906 		break;
907 #endif
908 
909 #ifdef NETATALK
910 	case ETHERTYPE_ATALK:
911 		isr = NETISR_ATALK;
912 		inq = &atintrq1;
913 		break;
914 
915 	case ETHERTYPE_AARP:
916 		aarpinput(ifp, m); /* XXX queue? */
917 		return;
918 #endif
919 
920 #ifdef MPLS
921 	case ETHERTYPE_MPLS:
922 		isr = NETISR_MPLS;
923 		inq = &mplsintrq;
924 		break;
925 #endif
926 
927 	default:
928 		goto drop;
929 	}
930 
931 	if (__predict_true(pktq)) {
932 #ifdef NET_MPSAFE
933 		const u_int h = curcpu()->ci_index;
934 #else
935 		const uint32_t h = pktq_rps_hash(m);
936 #endif
937 		if (__predict_false(!pktq_enqueue(pktq, m, h))) {
938 			m_freem(m);
939 		}
940 		return;
941 	}
942 
943 	if (__predict_false(!inq)) {
944 		/* Should not happen. */
945 		goto drop;
946 	}
947 
948 	IFQ_ENQUEUE_ISR(inq, m, isr);
949 	return;
950 
951 drop:
952 	m_freem(m);
953 dropped:
954 	if_statinc(ifp, if_ierrors); /* XXX should have a dedicated counter? */
955 }
956 
957 /*
958  * Convert Ethernet address to printable (loggable) representation.
959  */
960 char *
961 ether_sprintf(const u_char *ap)
962 {
963 	static char etherbuf[3 * ETHER_ADDR_LEN];
964 	return ether_snprintf(etherbuf, sizeof(etherbuf), ap);
965 }
966 
967 char *
968 ether_snprintf(char *buf, size_t len, const u_char *ap)
969 {
970 	char *cp = buf;
971 	size_t i;
972 
973 	for (i = 0; i < len / 3; i++) {
974 		*cp++ = hexdigits[*ap >> 4];
975 		*cp++ = hexdigits[*ap++ & 0xf];
976 		*cp++ = ':';
977 	}
978 	*--cp = '\0';
979 	return buf;
980 }
981 
982 /*
983  * Perform common duties while attaching to interface list
984  */
985 void
986 ether_ifattach(struct ifnet *ifp, const uint8_t *lla)
987 {
988 	struct ethercom *ec = (struct ethercom *)ifp;
989 
990 	ifp->if_type = IFT_ETHER;
991 	ifp->if_hdrlen = ETHER_HDR_LEN;
992 	ifp->if_dlt = DLT_EN10MB;
993 	ifp->if_mtu = ETHERMTU;
994 	ifp->if_output = ether_output;
995 	ifp->_if_input = ether_input;
996 	if (ifp->if_baudrate == 0)
997 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
998 
999 	if (lla != NULL)
1000 		if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla));
1001 
1002 	LIST_INIT(&ec->ec_multiaddrs);
1003 	SIMPLEQ_INIT(&ec->ec_vids);
1004 	ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
1005 	ec->ec_flags = 0;
1006 	ifp->if_broadcastaddr = etherbroadcastaddr;
1007 	bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header));
1008 #ifdef MBUFTRACE
1009 	mowner_init_owner(&ec->ec_tx_mowner, ifp->if_xname, "tx");
1010 	mowner_init_owner(&ec->ec_rx_mowner, ifp->if_xname, "rx");
1011 	MOWNER_ATTACH(&ec->ec_tx_mowner);
1012 	MOWNER_ATTACH(&ec->ec_rx_mowner);
1013 	ifp->if_mowner = &ec->ec_tx_mowner;
1014 #endif
1015 }
1016 
1017 void
1018 ether_ifdetach(struct ifnet *ifp)
1019 {
1020 	struct ethercom *ec = (void *) ifp;
1021 	struct ether_multi *enm;
1022 
1023 	IFNET_ASSERT_UNLOCKED(ifp);
1024 	/*
1025 	 * Prevent further calls to ioctl (for example turning off
1026 	 * promiscuous mode from the bridge code), which eventually can
1027 	 * call if_init() which can cause panics because the interface
1028 	 * is in the process of being detached. Return device not configured
1029 	 * instead.
1030 	 */
1031 	ifp->if_ioctl = __FPTRCAST(int (*)(struct ifnet *, u_long, void *),
1032 	    enxio);
1033 
1034 #if NBRIDGE > 0
1035 	if (ifp->if_bridge)
1036 		bridge_ifdetach(ifp);
1037 #endif
1038 	bpf_detach(ifp);
1039 #if NVLAN > 0
1040 	if (ec->ec_nvlans)
1041 		vlan_ifdetach(ifp);
1042 #endif
1043 
1044 	ETHER_LOCK(ec);
1045 	KASSERT(ec->ec_nvlans == 0);
1046 	while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) {
1047 		LIST_REMOVE(enm, enm_list);
1048 		kmem_free(enm, sizeof(*enm));
1049 		ec->ec_multicnt--;
1050 	}
1051 	ETHER_UNLOCK(ec);
1052 
1053 	mutex_obj_free(ec->ec_lock);
1054 	ec->ec_lock = NULL;
1055 
1056 	ifp->if_mowner = NULL;
1057 	MOWNER_DETACH(&ec->ec_rx_mowner);
1058 	MOWNER_DETACH(&ec->ec_tx_mowner);
1059 }
1060 
1061 #if 0
1062 /*
1063  * This is for reference.  We have a table-driven version
1064  * of the little-endian crc32 generator, which is faster
1065  * than the double-loop.
1066  */
1067 uint32_t
1068 ether_crc32_le(const uint8_t *buf, size_t len)
1069 {
1070 	uint32_t c, crc, carry;
1071 	size_t i, j;
1072 
1073 	crc = 0xffffffffU;	/* initial value */
1074 
1075 	for (i = 0; i < len; i++) {
1076 		c = buf[i];
1077 		for (j = 0; j < 8; j++) {
1078 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1079 			crc >>= 1;
1080 			c >>= 1;
1081 			if (carry)
1082 				crc = (crc ^ ETHER_CRC_POLY_LE);
1083 		}
1084 	}
1085 
1086 	return (crc);
1087 }
1088 #else
1089 uint32_t
1090 ether_crc32_le(const uint8_t *buf, size_t len)
1091 {
1092 	static const uint32_t crctab[] = {
1093 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1094 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1095 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1096 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1097 	};
1098 	uint32_t crc;
1099 	size_t i;
1100 
1101 	crc = 0xffffffffU;	/* initial value */
1102 
1103 	for (i = 0; i < len; i++) {
1104 		crc ^= buf[i];
1105 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1106 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1107 	}
1108 
1109 	return (crc);
1110 }
1111 #endif
1112 
1113 uint32_t
1114 ether_crc32_be(const uint8_t *buf, size_t len)
1115 {
1116 	uint32_t c, crc, carry;
1117 	size_t i, j;
1118 
1119 	crc = 0xffffffffU;	/* initial value */
1120 
1121 	for (i = 0; i < len; i++) {
1122 		c = buf[i];
1123 		for (j = 0; j < 8; j++) {
1124 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1125 			crc <<= 1;
1126 			c >>= 1;
1127 			if (carry)
1128 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1129 		}
1130 	}
1131 
1132 	return (crc);
1133 }
1134 
1135 #ifdef INET
1136 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] =
1137     { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 };
1138 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] =
1139     { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff };
1140 #endif
1141 #ifdef INET6
1142 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] =
1143     { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 };
1144 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] =
1145     { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff };
1146 #endif
1147 
1148 /*
1149  * ether_aton implementation, not using a static buffer.
1150  */
1151 int
1152 ether_aton_r(u_char *dest, size_t len, const char *str)
1153 {
1154 	const u_char *cp = (const void *)str;
1155 	u_char *ep;
1156 
1157 #define atox(c)	(((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10))
1158 
1159 	if (len < ETHER_ADDR_LEN)
1160 		return ENOSPC;
1161 
1162 	ep = dest + ETHER_ADDR_LEN;
1163 
1164 	while (*cp) {
1165 		if (!isxdigit(*cp))
1166 			return EINVAL;
1167 
1168 		*dest = atox(*cp);
1169 		cp++;
1170 		if (isxdigit(*cp)) {
1171 			*dest = (*dest << 4) | atox(*cp);
1172 			cp++;
1173 		}
1174 		dest++;
1175 
1176 		if (dest == ep)
1177 			return (*cp == '\0') ? 0 : ENAMETOOLONG;
1178 
1179 		switch (*cp) {
1180 		case ':':
1181 		case '-':
1182 		case '.':
1183 			cp++;
1184 			break;
1185 		}
1186 	}
1187 	return ENOBUFS;
1188 }
1189 
1190 /*
1191  * Convert a sockaddr into an Ethernet address or range of Ethernet
1192  * addresses.
1193  */
1194 int
1195 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN],
1196     uint8_t addrhi[ETHER_ADDR_LEN])
1197 {
1198 #ifdef INET
1199 	const struct sockaddr_in *sin;
1200 #endif
1201 #ifdef INET6
1202 	const struct sockaddr_in6 *sin6;
1203 #endif
1204 
1205 	switch (sa->sa_family) {
1206 
1207 	case AF_UNSPEC:
1208 		memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN);
1209 		memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1210 		break;
1211 
1212 #ifdef INET
1213 	case AF_INET:
1214 		sin = satocsin(sa);
1215 		if (sin->sin_addr.s_addr == INADDR_ANY) {
1216 			/*
1217 			 * An IP address of INADDR_ANY means listen to
1218 			 * or stop listening to all of the Ethernet
1219 			 * multicast addresses used for IP.
1220 			 * (This is for the sake of IP multicast routers.)
1221 			 */
1222 			memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN);
1223 			memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN);
1224 		} else {
1225 			ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo);
1226 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1227 		}
1228 		break;
1229 #endif
1230 #ifdef INET6
1231 	case AF_INET6:
1232 		sin6 = satocsin6(sa);
1233 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1234 			/*
1235 			 * An IP6 address of 0 means listen to or stop
1236 			 * listening to all of the Ethernet multicast
1237 			 * address used for IP6.
1238 			 * (This is used for multicast routers.)
1239 			 */
1240 			memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN);
1241 			memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN);
1242 		} else {
1243 			ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo);
1244 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1245 		}
1246 		break;
1247 #endif
1248 
1249 	default:
1250 		return EAFNOSUPPORT;
1251 	}
1252 	return 0;
1253 }
1254 
1255 /*
1256  * Add an Ethernet multicast address or range of addresses to the list for a
1257  * given interface.
1258  */
1259 int
1260 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec)
1261 {
1262 	struct ether_multi *enm, *_enm;
1263 	u_char addrlo[ETHER_ADDR_LEN];
1264 	u_char addrhi[ETHER_ADDR_LEN];
1265 	int error = 0;
1266 
1267 	/* Allocate out of lock */
1268 	enm = kmem_alloc(sizeof(*enm), KM_SLEEP);
1269 
1270 	ETHER_LOCK(ec);
1271 	error = ether_multiaddr(sa, addrlo, addrhi);
1272 	if (error != 0)
1273 		goto out;
1274 
1275 	/*
1276 	 * Verify that we have valid Ethernet multicast addresses.
1277 	 */
1278 	if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) {
1279 		error = EINVAL;
1280 		goto out;
1281 	}
1282 
1283 	/*
1284 	 * See if the address range is already in the list.
1285 	 */
1286 	_enm = ether_lookup_multi(addrlo, addrhi, ec);
1287 	if (_enm != NULL) {
1288 		/*
1289 		 * Found it; just increment the reference count.
1290 		 */
1291 		++_enm->enm_refcount;
1292 		error = 0;
1293 		goto out;
1294 	}
1295 
1296 	/*
1297 	 * Link a new multicast record into the interface's multicast list.
1298 	 */
1299 	memcpy(enm->enm_addrlo, addrlo, ETHER_ADDR_LEN);
1300 	memcpy(enm->enm_addrhi, addrhi, ETHER_ADDR_LEN);
1301 	enm->enm_refcount = 1;
1302 	LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list);
1303 	ec->ec_multicnt++;
1304 
1305 	/*
1306 	 * Return ENETRESET to inform the driver that the list has changed
1307 	 * and its reception filter should be adjusted accordingly.
1308 	 */
1309 	error = ENETRESET;
1310 	enm = NULL;
1311 
1312 out:
1313 	ETHER_UNLOCK(ec);
1314 	if (enm != NULL)
1315 		kmem_free(enm, sizeof(*enm));
1316 	return error;
1317 }
1318 
1319 /*
1320  * Delete a multicast address record.
1321  */
1322 int
1323 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec)
1324 {
1325 	struct ether_multi *enm;
1326 	u_char addrlo[ETHER_ADDR_LEN];
1327 	u_char addrhi[ETHER_ADDR_LEN];
1328 	int error;
1329 
1330 	ETHER_LOCK(ec);
1331 	error = ether_multiaddr(sa, addrlo, addrhi);
1332 	if (error != 0)
1333 		goto error;
1334 
1335 	/*
1336 	 * Look up the address in our list.
1337 	 */
1338 	enm = ether_lookup_multi(addrlo, addrhi, ec);
1339 	if (enm == NULL) {
1340 		error = ENXIO;
1341 		goto error;
1342 	}
1343 	if (--enm->enm_refcount != 0) {
1344 		/*
1345 		 * Still some claims to this record.
1346 		 */
1347 		error = 0;
1348 		goto error;
1349 	}
1350 
1351 	/*
1352 	 * No remaining claims to this record; unlink and free it.
1353 	 */
1354 	LIST_REMOVE(enm, enm_list);
1355 	ec->ec_multicnt--;
1356 	ETHER_UNLOCK(ec);
1357 	kmem_free(enm, sizeof(*enm));
1358 
1359 	/*
1360 	 * Return ENETRESET to inform the driver that the list has changed
1361 	 * and its reception filter should be adjusted accordingly.
1362 	 */
1363 	return ENETRESET;
1364 
1365 error:
1366 	ETHER_UNLOCK(ec);
1367 	return error;
1368 }
1369 
1370 void
1371 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb)
1372 {
1373 	ec->ec_ifflags_cb = cb;
1374 }
1375 
1376 void
1377 ether_set_vlan_cb(struct ethercom *ec, ether_vlancb_t cb)
1378 {
1379 
1380 	ec->ec_vlan_cb = cb;
1381 }
1382 
1383 static int
1384 ether_ioctl_reinit(struct ethercom *ec)
1385 {
1386 	struct ifnet *ifp = &ec->ec_if;
1387 	int error;
1388 
1389 	switch (ifp->if_flags & (IFF_UP | IFF_RUNNING)) {
1390 	case IFF_RUNNING:
1391 		/*
1392 		 * If interface is marked down and it is running,
1393 		 * then stop and disable it.
1394 		 */
1395 		(*ifp->if_stop)(ifp, 1);
1396 		break;
1397 	case IFF_UP:
1398 		/*
1399 		 * If interface is marked up and it is stopped, then
1400 		 * start it.
1401 		 */
1402 		return (*ifp->if_init)(ifp);
1403 	case IFF_UP | IFF_RUNNING:
1404 		error = 0;
1405 		if (ec->ec_ifflags_cb != NULL) {
1406 			error = (*ec->ec_ifflags_cb)(ec);
1407 			if (error == ENETRESET) {
1408 				/*
1409 				 * Reset the interface to pick up
1410 				 * changes in any other flags that
1411 				 * affect the hardware state.
1412 				 */
1413 				return (*ifp->if_init)(ifp);
1414 			}
1415 		} else
1416 			error = (*ifp->if_init)(ifp);
1417 		return error;
1418 	case 0:
1419 		break;
1420 	}
1421 
1422 	return 0;
1423 }
1424 
1425 /*
1426  * Common ioctls for Ethernet interfaces.  Note, we must be
1427  * called at splnet().
1428  */
1429 int
1430 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data)
1431 {
1432 	struct ethercom *ec = (void *)ifp;
1433 	struct eccapreq *eccr;
1434 	struct ifreq *ifr = (struct ifreq *)data;
1435 	struct if_laddrreq *iflr = data;
1436 	const struct sockaddr_dl *sdl;
1437 	static const uint8_t zero[ETHER_ADDR_LEN];
1438 	int error;
1439 
1440 	switch (cmd) {
1441 	case SIOCINITIFADDR:
1442 	    {
1443 		struct ifaddr *ifa = (struct ifaddr *)data;
1444 		if (ifa->ifa_addr->sa_family != AF_LINK
1445 		    && (ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
1446 		       (IFF_UP | IFF_RUNNING)) {
1447 			ifp->if_flags |= IFF_UP;
1448 			if ((error = (*ifp->if_init)(ifp)) != 0)
1449 				return error;
1450 		}
1451 #ifdef INET
1452 		if (ifa->ifa_addr->sa_family == AF_INET)
1453 			arp_ifinit(ifp, ifa);
1454 #endif
1455 		return 0;
1456 	    }
1457 
1458 	case SIOCSIFMTU:
1459 	    {
1460 		int maxmtu;
1461 
1462 		if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU)
1463 			maxmtu = ETHERMTU_JUMBO;
1464 		else
1465 			maxmtu = ETHERMTU;
1466 
1467 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu)
1468 			return EINVAL;
1469 		else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET)
1470 			return error;
1471 		else if (ifp->if_flags & IFF_UP) {
1472 			/* Make sure the device notices the MTU change. */
1473 			return (*ifp->if_init)(ifp);
1474 		} else
1475 			return 0;
1476 	    }
1477 
1478 	case SIOCSIFFLAGS:
1479 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
1480 			return error;
1481 		return ether_ioctl_reinit(ec);
1482 	case SIOCGIFFLAGS:
1483 		error = ifioctl_common(ifp, cmd, data);
1484 		if (error == 0) {
1485 			/* Set IFF_ALLMULTI for backcompat */
1486 			ifr->ifr_flags |= (ec->ec_flags & ETHER_F_ALLMULTI) ?
1487 			    IFF_ALLMULTI : 0;
1488 		}
1489 		return error;
1490 	case SIOCGETHERCAP:
1491 		eccr = (struct eccapreq *)data;
1492 		eccr->eccr_capabilities = ec->ec_capabilities;
1493 		eccr->eccr_capenable = ec->ec_capenable;
1494 		return 0;
1495 	case SIOCSETHERCAP:
1496 		eccr = (struct eccapreq *)data;
1497 		if ((eccr->eccr_capenable & ~ec->ec_capabilities) != 0)
1498 			return EINVAL;
1499 		if (eccr->eccr_capenable == ec->ec_capenable)
1500 			return 0;
1501 #if 0 /* notyet */
1502 		ec->ec_capenable = (ec->ec_capenable & ETHERCAP_CANTCHANGE)
1503 		    | (eccr->eccr_capenable & ~ETHERCAP_CANTCHANGE);
1504 #else
1505 		ec->ec_capenable = eccr->eccr_capenable;
1506 #endif
1507 		return ether_ioctl_reinit(ec);
1508 	case SIOCADDMULTI:
1509 		return ether_addmulti(ifreq_getaddr(cmd, ifr), ec);
1510 	case SIOCDELMULTI:
1511 		return ether_delmulti(ifreq_getaddr(cmd, ifr), ec);
1512 	case SIOCSIFMEDIA:
1513 	case SIOCGIFMEDIA:
1514 		if (ec->ec_mii != NULL)
1515 			return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media,
1516 			    cmd);
1517 		else if (ec->ec_ifmedia != NULL)
1518 			return ifmedia_ioctl(ifp, ifr, ec->ec_ifmedia, cmd);
1519 		else
1520 			return ENOTTY;
1521 		break;
1522 	case SIOCALIFADDR:
1523 		sdl = satocsdl(sstocsa(&iflr->addr));
1524 		if (sdl->sdl_family != AF_LINK)
1525 			;
1526 		else if (ETHER_IS_MULTICAST(CLLADDR(sdl)))
1527 			return EINVAL;
1528 		else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0)
1529 			return EINVAL;
1530 		/*FALLTHROUGH*/
1531 	default:
1532 		return ifioctl_common(ifp, cmd, data);
1533 	}
1534 	return 0;
1535 }
1536 
1537 /*
1538  * Enable/disable passing VLAN packets if the parent interface supports it.
1539  * Return:
1540  * 	 0: Ok
1541  *	-1: Parent interface does not support vlans
1542  *	>0: Error
1543  */
1544 int
1545 ether_enable_vlan_mtu(struct ifnet *ifp)
1546 {
1547 	int error;
1548 	struct ethercom *ec = (void *)ifp;
1549 
1550 	/* Parent does not support VLAN's */
1551 	if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0)
1552 		return -1;
1553 
1554 	/*
1555 	 * Parent supports the VLAN_MTU capability,
1556 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames;
1557 	 * enable it.
1558 	 */
1559 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1560 
1561 	/* Interface is down, defer for later */
1562 	if ((ifp->if_flags & IFF_UP) == 0)
1563 		return 0;
1564 
1565 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1566 		return 0;
1567 
1568 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1569 	return error;
1570 }
1571 
1572 int
1573 ether_disable_vlan_mtu(struct ifnet *ifp)
1574 {
1575 	int error;
1576 	struct ethercom *ec = (void *)ifp;
1577 
1578 	/* We still have VLAN's, defer for later */
1579 	if (ec->ec_nvlans != 0)
1580 		return 0;
1581 
1582 	/* Parent does not support VLAB's, nothing to do. */
1583 	if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0)
1584 		return -1;
1585 
1586 	/*
1587 	 * Disable Tx/Rx of VLAN-sized frames.
1588 	 */
1589 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1590 
1591 	/* Interface is down, defer for later */
1592 	if ((ifp->if_flags & IFF_UP) == 0)
1593 		return 0;
1594 
1595 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1596 		return 0;
1597 
1598 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1599 	return error;
1600 }
1601 
1602 static int
1603 ether_multicast_sysctl(SYSCTLFN_ARGS)
1604 {
1605 	struct ether_multi *enm;
1606 	struct ifnet *ifp;
1607 	struct ethercom *ec;
1608 	int error = 0;
1609 	size_t written;
1610 	struct psref psref;
1611 	int bound;
1612 	unsigned int multicnt;
1613 	struct ether_multi_sysctl *addrs;
1614 	int i;
1615 
1616 	if (namelen != 1)
1617 		return EINVAL;
1618 
1619 	bound = curlwp_bind();
1620 	ifp = if_get_byindex(name[0], &psref);
1621 	if (ifp == NULL) {
1622 		error = ENODEV;
1623 		goto out;
1624 	}
1625 	if (ifp->if_type != IFT_ETHER) {
1626 		if_put(ifp, &psref);
1627 		*oldlenp = 0;
1628 		goto out;
1629 	}
1630 	ec = (struct ethercom *)ifp;
1631 
1632 	if (oldp == NULL) {
1633 		if_put(ifp, &psref);
1634 		*oldlenp = ec->ec_multicnt * sizeof(*addrs);
1635 		goto out;
1636 	}
1637 
1638 	/*
1639 	 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which
1640 	 * is sleepable, while holding it. Copy data to a local buffer first
1641 	 * with the lock taken and then call sysctl_copyout without holding it.
1642 	 */
1643 retry:
1644 	multicnt = ec->ec_multicnt;
1645 
1646 	if (multicnt == 0) {
1647 		if_put(ifp, &psref);
1648 		*oldlenp = 0;
1649 		goto out;
1650 	}
1651 
1652 	addrs = kmem_zalloc(sizeof(*addrs) * multicnt, KM_SLEEP);
1653 
1654 	ETHER_LOCK(ec);
1655 	if (multicnt != ec->ec_multicnt) {
1656 		/* The number of multicast addresses has changed */
1657 		ETHER_UNLOCK(ec);
1658 		kmem_free(addrs, sizeof(*addrs) * multicnt);
1659 		goto retry;
1660 	}
1661 
1662 	i = 0;
1663 	LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) {
1664 		struct ether_multi_sysctl *addr = &addrs[i];
1665 		addr->enm_refcount = enm->enm_refcount;
1666 		memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN);
1667 		memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN);
1668 		i++;
1669 	}
1670 	ETHER_UNLOCK(ec);
1671 
1672 	error = 0;
1673 	written = 0;
1674 	for (i = 0; i < multicnt; i++) {
1675 		struct ether_multi_sysctl *addr = &addrs[i];
1676 
1677 		if (written + sizeof(*addr) > *oldlenp)
1678 			break;
1679 		error = sysctl_copyout(l, addr, oldp, sizeof(*addr));
1680 		if (error)
1681 			break;
1682 		written += sizeof(*addr);
1683 		oldp = (char *)oldp + sizeof(*addr);
1684 	}
1685 	kmem_free(addrs, sizeof(*addrs) * multicnt);
1686 
1687 	if_put(ifp, &psref);
1688 
1689 	*oldlenp = written;
1690 out:
1691 	curlwp_bindx(bound);
1692 	return error;
1693 }
1694 
1695 static void
1696 ether_sysctl_setup(struct sysctllog **clog)
1697 {
1698 	const struct sysctlnode *rnode = NULL;
1699 
1700 	sysctl_createv(clog, 0, NULL, &rnode,
1701 		       CTLFLAG_PERMANENT,
1702 		       CTLTYPE_NODE, "ether",
1703 		       SYSCTL_DESCR("Ethernet-specific information"),
1704 		       NULL, 0, NULL, 0,
1705 		       CTL_NET, CTL_CREATE, CTL_EOL);
1706 
1707 	sysctl_createv(clog, 0, &rnode, NULL,
1708 		       CTLFLAG_PERMANENT,
1709 		       CTLTYPE_NODE, "multicast",
1710 		       SYSCTL_DESCR("multicast addresses"),
1711 		       ether_multicast_sysctl, 0, NULL, 0,
1712 		       CTL_CREATE, CTL_EOL);
1713 }
1714 
1715 void
1716 etherinit(void)
1717 {
1718 
1719 #ifdef DIAGNOSTIC
1720 	mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET);
1721 #endif
1722 	ether_sysctl_setup(NULL);
1723 }
1724