xref: /netbsd-src/sys/net/if_ethersubr.c (revision 627f7eb200a4419d89b531d55fccd2ee3ffdcde0)
1 /*	$NetBSD: if_ethersubr.c,v 1.292 2021/02/14 19:35:37 roy 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 1.292 2021/02/14 19:35:37 roy 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 CTASSERT(sizeof(struct ether_addr) == 6);
168 CTASSERT(sizeof(struct ether_header) == 14);
169 
170 #ifdef DIAGNOSTIC
171 static struct timeval bigpktppslim_last;
172 static int bigpktppslim = 2;	/* XXX */
173 static int bigpktpps_count;
174 static kmutex_t bigpktpps_lock __cacheline_aligned;
175 #endif
176 
177 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] =
178     { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
179 const uint8_t ethermulticastaddr_slowprotocols[ETHER_ADDR_LEN] =
180     { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x02 };
181 #define senderr(e) { error = (e); goto bad;}
182 
183 static int ether_output(struct ifnet *, struct mbuf *,
184     const struct sockaddr *, const struct rtentry *);
185 
186 /*
187  * Ethernet output routine.
188  * Encapsulate a packet of type family for the local net.
189  * Assumes that ifp is actually pointer to ethercom structure.
190  */
191 static int
192 ether_output(struct ifnet * const ifp0, struct mbuf * const m0,
193     const struct sockaddr * const dst, const struct rtentry *rt)
194 {
195 	uint8_t esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
196 	uint16_t etype = 0;
197 	int error = 0, hdrcmplt = 0;
198 	struct mbuf *m = m0;
199 	struct mbuf *mcopy = NULL;
200 	struct ether_header *eh;
201 	struct ifnet *ifp = ifp0;
202 #ifdef INET
203 	struct arphdr *ah;
204 #endif
205 #ifdef NETATALK
206 	struct at_ifaddr *aa;
207 #endif
208 
209 #ifdef MBUFTRACE
210 	m_claimm(m, ifp->if_mowner);
211 #endif
212 
213 #if NCARP > 0
214 	if (ifp->if_type == IFT_CARP) {
215 		struct ifaddr *ifa;
216 		int s = pserialize_read_enter();
217 
218 		/* loop back if this is going to the carp interface */
219 		if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP &&
220 		    (ifa = ifa_ifwithaddr(dst)) != NULL) {
221 			if (ifa->ifa_ifp == ifp0) {
222 				pserialize_read_exit(s);
223 				return looutput(ifp0, m, dst, rt);
224 			}
225 		}
226 		pserialize_read_exit(s);
227 
228 		ifp = ifp->if_carpdev;
229 		/* ac = (struct arpcom *)ifp; */
230 
231 		if ((ifp0->if_flags & (IFF_UP | IFF_RUNNING)) !=
232 		    (IFF_UP | IFF_RUNNING))
233 			senderr(ENETDOWN);
234 	}
235 #endif
236 
237 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
238 		senderr(ENETDOWN);
239 
240 	switch (dst->sa_family) {
241 
242 #ifdef INET
243 	case AF_INET:
244 		if (m->m_flags & M_BCAST) {
245 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
246 		} else if (m->m_flags & M_MCAST) {
247 			ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst);
248 		} else {
249 			error = arpresolve(ifp0, rt, m, dst, edst, sizeof(edst));
250 			if (error)
251 				return (error == EWOULDBLOCK) ? 0 : error;
252 		}
253 		/* If broadcasting on a simplex interface, loopback a copy */
254 		if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
255 			mcopy = m_copypacket(m, M_DONTWAIT);
256 		etype = htons(ETHERTYPE_IP);
257 		break;
258 
259 	case AF_ARP:
260 		ah = mtod(m, struct arphdr *);
261 		if (m->m_flags & M_BCAST) {
262 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
263 		} else {
264 			void *tha = ar_tha(ah);
265 
266 			if (tha == NULL) {
267 				/* fake with ARPHRD_IEEE1394 */
268 				m_freem(m);
269 				return 0;
270 			}
271 			memcpy(edst, tha, sizeof(edst));
272 		}
273 
274 		ah->ar_hrd = htons(ARPHRD_ETHER);
275 
276 		switch (ntohs(ah->ar_op)) {
277 		case ARPOP_REVREQUEST:
278 		case ARPOP_REVREPLY:
279 			etype = htons(ETHERTYPE_REVARP);
280 			break;
281 
282 		case ARPOP_REQUEST:
283 		case ARPOP_REPLY:
284 		default:
285 			etype = htons(ETHERTYPE_ARP);
286 		}
287 		break;
288 #endif
289 
290 #ifdef INET6
291 	case AF_INET6:
292 		if (m->m_flags & M_BCAST) {
293 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
294 		} else if (m->m_flags & M_MCAST) {
295 			ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr,
296 			    edst);
297 		} else {
298 			error = nd6_resolve(ifp0, rt, m, dst, edst,
299 			    sizeof(edst));
300 			if (error)
301 				return (error == EWOULDBLOCK) ? 0 : error;
302 		}
303 		etype = htons(ETHERTYPE_IPV6);
304 		break;
305 #endif
306 
307 #ifdef NETATALK
308 	case AF_APPLETALK: {
309 		struct ifaddr *ifa;
310 		int s;
311 
312 		KERNEL_LOCK(1, NULL);
313 
314 		if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) {
315 			KERNEL_UNLOCK_ONE(NULL);
316 			return 0;
317 		}
318 
319 		/*
320 		 * ifaddr is the first thing in at_ifaddr
321 		 */
322 		s = pserialize_read_enter();
323 		ifa = at_ifawithnet((const struct sockaddr_at *)dst, ifp);
324 		if (ifa == NULL) {
325 			pserialize_read_exit(s);
326 			KERNEL_UNLOCK_ONE(NULL);
327 			senderr(EADDRNOTAVAIL);
328 		}
329 		aa = (struct at_ifaddr *)ifa;
330 
331 		/*
332 		 * In the phase 2 case, we need to prepend an mbuf for the
333 		 * llc header.
334 		 */
335 		if (aa->aa_flags & AFA_PHASE2) {
336 			struct llc llc;
337 
338 			M_PREPEND(m, sizeof(struct llc), M_DONTWAIT);
339 			if (m == NULL) {
340 				pserialize_read_exit(s);
341 				KERNEL_UNLOCK_ONE(NULL);
342 				senderr(ENOBUFS);
343 			}
344 
345 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
346 			llc.llc_control = LLC_UI;
347 			memcpy(llc.llc_snap_org_code, at_org_code,
348 			    sizeof(llc.llc_snap_org_code));
349 			llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK);
350 			memcpy(mtod(m, void *), &llc, sizeof(struct llc));
351 		} else {
352 			etype = htons(ETHERTYPE_ATALK);
353 		}
354 		pserialize_read_exit(s);
355 		KERNEL_UNLOCK_ONE(NULL);
356 		break;
357 	}
358 #endif /* NETATALK */
359 
360 	case pseudo_AF_HDRCMPLT:
361 		hdrcmplt = 1;
362 		memcpy(esrc,
363 		    ((const struct ether_header *)dst->sa_data)->ether_shost,
364 		    sizeof(esrc));
365 		/* FALLTHROUGH */
366 
367 	case AF_UNSPEC:
368 		memcpy(edst,
369 		    ((const struct ether_header *)dst->sa_data)->ether_dhost,
370 		    sizeof(edst));
371 		/* AF_UNSPEC doesn't swap the byte order of the ether_type. */
372 		etype = ((const struct ether_header *)dst->sa_data)->ether_type;
373 		break;
374 
375 	default:
376 		printf("%s: can't handle af%d\n", ifp->if_xname,
377 		    dst->sa_family);
378 		senderr(EAFNOSUPPORT);
379 	}
380 
381 #ifdef MPLS
382 	{
383 		struct m_tag *mtag;
384 		mtag = m_tag_find(m, PACKET_TAG_MPLS);
385 		if (mtag != NULL) {
386 			/* Having the tag itself indicates it's MPLS */
387 			etype = htons(ETHERTYPE_MPLS);
388 			m_tag_delete(m, mtag);
389 		}
390 	}
391 #endif
392 
393 	if (mcopy)
394 		(void)looutput(ifp, mcopy, dst, rt);
395 
396 	KASSERT((m->m_flags & M_PKTHDR) != 0);
397 
398 	/*
399 	 * If no ether type is set, this must be a 802.2 formatted packet.
400 	 */
401 	if (etype == 0)
402 		etype = htons(m->m_pkthdr.len);
403 
404 	/*
405 	 * Add local net header. If no space in first mbuf, allocate another.
406 	 */
407 	M_PREPEND(m, sizeof(struct ether_header), M_DONTWAIT);
408 	if (m == NULL)
409 		senderr(ENOBUFS);
410 
411 	eh = mtod(m, struct ether_header *);
412 	/* Note: etype is already in network byte order. */
413 	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 
422 #if NCARP > 0
423 	if (ifp0 != ifp && ifp0->if_type == IFT_CARP) {
424 	 	memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl),
425 		    sizeof(eh->ether_shost));
426 	}
427 #endif
428 
429 	if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0)
430 		return error;
431 	if (m == NULL)
432 		return 0;
433 
434 #if NBRIDGE > 0
435 	/*
436 	 * Bridges require special output handling.
437 	 */
438 	if (ifp->if_bridge)
439 		return bridge_output(ifp, m, NULL, NULL);
440 #endif
441 
442 #if NCARP > 0
443 	if (ifp != ifp0)
444 		if_statadd(ifp0, if_obytes, m->m_pkthdr.len + ETHER_HDR_LEN);
445 #endif
446 
447 #ifdef ALTQ
448 	KERNEL_LOCK(1, NULL);
449 	/*
450 	 * If ALTQ is enabled on the parent interface, do
451 	 * classification; the queueing discipline might not
452 	 * require classification, but might require the
453 	 * address family/header pointer in the pktattr.
454 	 */
455 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
456 		altq_etherclassify(&ifp->if_snd, m);
457 	KERNEL_UNLOCK_ONE(NULL);
458 #endif
459 	return ifq_enqueue(ifp, m);
460 
461 bad:
462 	if_statinc(ifp, if_oerrors);
463 	if (m)
464 		m_freem(m);
465 	return error;
466 }
467 
468 #ifdef ALTQ
469 /*
470  * This routine is a slight hack to allow a packet to be classified
471  * if the Ethernet headers are present.  It will go away when ALTQ's
472  * classification engine understands link headers.
473  *
474  * XXX: We may need to do m_pullups here. First to ensure struct ether_header
475  * is indeed contiguous, then to read the LLC and so on.
476  */
477 void
478 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m)
479 {
480 	struct ether_header *eh;
481 	struct mbuf *mtop = m;
482 	uint16_t ether_type;
483 	int hlen, af, hdrsize;
484 	void *hdr;
485 
486 	KASSERT((mtop->m_flags & M_PKTHDR) != 0);
487 
488 	hlen = ETHER_HDR_LEN;
489 	eh = mtod(m, struct ether_header *);
490 
491 	ether_type = htons(eh->ether_type);
492 
493 	if (ether_type < ETHERMTU) {
494 		/* LLC/SNAP */
495 		struct llc *llc = (struct llc *)(eh + 1);
496 		hlen += 8;
497 
498 		if (m->m_len < hlen ||
499 		    llc->llc_dsap != LLC_SNAP_LSAP ||
500 		    llc->llc_ssap != LLC_SNAP_LSAP ||
501 		    llc->llc_control != LLC_UI) {
502 			/* Not SNAP. */
503 			goto bad;
504 		}
505 
506 		ether_type = htons(llc->llc_un.type_snap.ether_type);
507 	}
508 
509 	switch (ether_type) {
510 	case ETHERTYPE_IP:
511 		af = AF_INET;
512 		hdrsize = 20;		/* sizeof(struct ip) */
513 		break;
514 
515 	case ETHERTYPE_IPV6:
516 		af = AF_INET6;
517 		hdrsize = 40;		/* sizeof(struct ip6_hdr) */
518 		break;
519 
520 	default:
521 		af = AF_UNSPEC;
522 		hdrsize = 0;
523 		break;
524 	}
525 
526 	while (m->m_len <= hlen) {
527 		hlen -= m->m_len;
528 		m = m->m_next;
529 		if (m == NULL)
530 			goto bad;
531 	}
532 
533 	if (m->m_len < (hlen + hdrsize)) {
534 		/*
535 		 * protocol header not in a single mbuf.
536 		 * We can't cope with this situation right
537 		 * now (but it shouldn't ever happen, really, anyhow).
538 		 */
539 #ifdef DEBUG
540 		printf("altq_etherclassify: headers span multiple mbufs: "
541 		    "%d < %d\n", m->m_len, (hlen + hdrsize));
542 #endif
543 		goto bad;
544 	}
545 
546 	m->m_data += hlen;
547 	m->m_len -= hlen;
548 
549 	hdr = mtod(m, void *);
550 
551 	if (ALTQ_NEEDS_CLASSIFY(ifq)) {
552 		mtop->m_pkthdr.pattr_class =
553 		    (*ifq->altq_classify)(ifq->altq_clfier, m, af);
554 	}
555 	mtop->m_pkthdr.pattr_af = af;
556 	mtop->m_pkthdr.pattr_hdr = hdr;
557 
558 	m->m_data -= hlen;
559 	m->m_len += hlen;
560 
561 	return;
562 
563 bad:
564 	mtop->m_pkthdr.pattr_class = NULL;
565 	mtop->m_pkthdr.pattr_hdr = NULL;
566 	mtop->m_pkthdr.pattr_af = AF_UNSPEC;
567 }
568 #endif /* ALTQ */
569 
570 #if defined (LLC) || defined (NETATALK)
571 static void
572 ether_input_llc(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh)
573 {
574 	struct ifqueue *inq = NULL;
575 	int isr = 0;
576 	struct llc *l;
577 
578 	if (m->m_len < sizeof(*eh) + sizeof(struct llc))
579 		goto drop;
580 
581 	l = (struct llc *)(eh+1);
582 	switch (l->llc_dsap) {
583 #ifdef NETATALK
584 	case LLC_SNAP_LSAP:
585 		switch (l->llc_control) {
586 		case LLC_UI:
587 			if (l->llc_ssap != LLC_SNAP_LSAP)
588 				goto drop;
589 
590 			if (memcmp(&(l->llc_snap_org_code)[0],
591 			    at_org_code, sizeof(at_org_code)) == 0 &&
592 			    ntohs(l->llc_snap_ether_type) ==
593 			    ETHERTYPE_ATALK) {
594 				inq = &atintrq2;
595 				m_adj(m, sizeof(struct ether_header)
596 				    + sizeof(struct llc));
597 				isr = NETISR_ATALK;
598 				break;
599 			}
600 
601 			if (memcmp(&(l->llc_snap_org_code)[0],
602 			    aarp_org_code,
603 			    sizeof(aarp_org_code)) == 0 &&
604 			    ntohs(l->llc_snap_ether_type) ==
605 			    ETHERTYPE_AARP) {
606 				m_adj(m, sizeof(struct ether_header)
607 				    + sizeof(struct llc));
608 				aarpinput(ifp, m); /* XXX queue? */
609 				return;
610 			}
611 
612 		default:
613 			goto drop;
614 		}
615 		break;
616 #endif
617 	default:
618 		goto drop;
619 	}
620 
621 	KASSERT(inq != NULL);
622 	IFQ_ENQUEUE_ISR(inq, m, isr);
623 	return;
624 
625 drop:
626 	m_freem(m);
627 	if_statinc(ifp, if_ierrors); /* XXX should have a dedicated counter? */
628 	return;
629 }
630 #endif /* defined (LLC) || defined (NETATALK) */
631 
632 /*
633  * Process a received Ethernet packet;
634  * the packet is in the mbuf chain m with
635  * the ether header.
636  */
637 void
638 ether_input(struct ifnet *ifp, struct mbuf *m)
639 {
640 	struct ethercom *ec = (struct ethercom *) ifp;
641 	pktqueue_t *pktq = NULL;
642 	struct ifqueue *inq = NULL;
643 	uint16_t etype;
644 	struct ether_header *eh;
645 	size_t ehlen;
646 	static int earlypkts;
647 	int isr = 0;
648 
649 	KASSERT(!cpu_intr_p());
650 	KASSERT((m->m_flags & M_PKTHDR) != 0);
651 
652 	if ((ifp->if_flags & IFF_UP) == 0)
653 		goto drop;
654 
655 #ifdef MBUFTRACE
656 	m_claimm(m, &ec->ec_rx_mowner);
657 #endif
658 
659 	if (__predict_false(m->m_len < sizeof(*eh))) {
660 		if ((m = m_pullup(m, sizeof(*eh))) == NULL)
661 			goto dropped;
662 	}
663 
664 	eh = mtod(m, struct ether_header *);
665 	etype = ntohs(eh->ether_type);
666 	ehlen = sizeof(*eh);
667 
668 	if (__predict_false(earlypkts < 100 ||
669 		entropy_epoch() == (unsigned)-1)) {
670 		rnd_add_data(NULL, eh, ehlen, 0);
671 		earlypkts++;
672 	}
673 
674 	/*
675 	 * Determine if the packet is within its size limits. For MPLS the
676 	 * header length is variable, so we skip the check.
677 	 */
678 	if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len >
679 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
680 #ifdef DIAGNOSTIC
681 		mutex_enter(&bigpktpps_lock);
682 		if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count,
683 		    bigpktppslim)) {
684 			printf("%s: discarding oversize frame (len=%d)\n",
685 			    ifp->if_xname, m->m_pkthdr.len);
686 		}
687 		mutex_exit(&bigpktpps_lock);
688 #endif
689 		goto drop;
690 	}
691 
692 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
693 		/*
694 		 * If this is not a simplex interface, drop the packet
695 		 * if it came from us.
696 		 */
697 		if ((ifp->if_flags & IFF_SIMPLEX) == 0 &&
698 		    memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost,
699 		    ETHER_ADDR_LEN) == 0) {
700 			goto drop;
701 		}
702 
703 		if (memcmp(etherbroadcastaddr,
704 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
705 			m->m_flags |= M_BCAST;
706 		else
707 			m->m_flags |= M_MCAST;
708 		if_statinc(ifp, if_imcasts);
709 	}
710 
711 	/* If the CRC is still on the packet, trim it off. */
712 	if (m->m_flags & M_HASFCS) {
713 		m_adj(m, -ETHER_CRC_LEN);
714 		m->m_flags &= ~M_HASFCS;
715 	}
716 
717 	if_statadd(ifp, if_ibytes, m->m_pkthdr.len);
718 
719 #if NCARP > 0
720 	if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) {
721 		/*
722 		 * Clear M_PROMISC, in case the packet comes from a
723 		 * vlan.
724 		 */
725 		m->m_flags &= ~M_PROMISC;
726 		if (carp_input(m, (uint8_t *)&eh->ether_shost,
727 		    (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0)
728 			return;
729 	}
730 #endif
731 
732 	if ((m->m_flags & (M_BCAST | M_MCAST | M_PROMISC)) == 0 &&
733 	    (ifp->if_flags & IFF_PROMISC) != 0 &&
734 	    memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost,
735 	     ETHER_ADDR_LEN) != 0) {
736 		m->m_flags |= M_PROMISC;
737 	}
738 
739 	if ((m->m_flags & M_PROMISC) == 0) {
740 		if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
741 			return;
742 		if (m == NULL)
743 			return;
744 
745 		eh = mtod(m, struct ether_header *);
746 		etype = ntohs(eh->ether_type);
747 	}
748 
749 #if NAGR > 0
750 	if (ifp->if_agrprivate &&
751 	    __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) {
752 		m->m_flags &= ~M_PROMISC;
753 		agr_input(ifp, m);
754 		return;
755 	}
756 #endif
757 
758 	/*
759 	 * If VLANs are configured on the interface, check to
760 	 * see if the device performed the decapsulation and
761 	 * provided us with the tag.
762 	 */
763 	if (ec->ec_nvlans && vlan_has_tag(m)) {
764 #if NVLAN > 0
765 		/*
766 		 * vlan_input() will either recursively call ether_input()
767 		 * or drop the packet.
768 		 */
769 		vlan_input(ifp, m);
770 		return;
771 #else
772 		goto drop;
773 #endif
774 	}
775 
776 	/*
777 	 * Handle protocols that expect to have the Ethernet header
778 	 * (and possibly FCS) intact.
779 	 */
780 	switch (etype) {
781 	case ETHERTYPE_VLAN: {
782 		struct ether_vlan_header *evl = (void *)eh;
783 
784 		/*
785 		 * If there is a tag of 0, then the VLAN header was probably
786 		 * just being used to store the priority.  Extract the ether
787 		 * type, and if IP or IPV6, let them deal with it.
788 		 */
789 		if (m->m_len >= sizeof(*evl) &&
790 		    EVL_VLANOFTAG(evl->evl_tag) == 0) {
791 			etype = ntohs(evl->evl_proto);
792 			ehlen = sizeof(*evl);
793 			if ((m->m_flags & M_PROMISC) == 0 &&
794 			    (etype == ETHERTYPE_IP ||
795 			     etype == ETHERTYPE_IPV6))
796 				break;
797 		}
798 
799 #if NVLAN > 0
800 		/*
801 		 * vlan_input() will either recursively call ether_input()
802 		 * or drop the packet.
803 		 */
804 		if (ec->ec_nvlans != 0) {
805 			vlan_input(ifp, m);
806 			return;
807 		} else
808 #endif
809 			goto drop;
810 	}
811 
812 #if NPPPOE > 0
813 	case ETHERTYPE_PPPOEDISC:
814 		pppoedisc_input(ifp, m);
815 		return;
816 
817 	case ETHERTYPE_PPPOE:
818 		pppoe_input(ifp, m);
819 		return;
820 #endif
821 
822 	case ETHERTYPE_SLOWPROTOCOLS: {
823 		uint8_t subtype;
824 
825 		if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype))
826 			goto drop;
827 
828 		m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype);
829 		switch (subtype) {
830 #if NAGR > 0
831 		case SLOWPROTOCOLS_SUBTYPE_LACP:
832 			if (ifp->if_agrprivate) {
833 				ieee8023ad_lacp_input(ifp, m);
834 				return;
835 			}
836 			break;
837 
838 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
839 			if (ifp->if_agrprivate) {
840 				ieee8023ad_marker_input(ifp, m);
841 				return;
842 			}
843 			break;
844 #endif
845 
846 		default:
847 			if (subtype == 0 || subtype > 10) {
848 				/* illegal value */
849 				goto drop;
850 			}
851 			/* unknown subtype */
852 			break;
853 		}
854 	}
855 	/* FALLTHROUGH */
856 	default:
857 		if (m->m_flags & M_PROMISC)
858 			goto drop;
859 	}
860 
861 	/* If the CRC is still on the packet, trim it off. */
862 	if (m->m_flags & M_HASFCS) {
863 		m_adj(m, -ETHER_CRC_LEN);
864 		m->m_flags &= ~M_HASFCS;
865 	}
866 
867 	/* etype represents the size of the payload in this case */
868 	if (etype <= ETHERMTU + sizeof(struct ether_header)) {
869 		KASSERT(ehlen == sizeof(*eh));
870 #if defined (LLC) || defined (NETATALK)
871 		ether_input_llc(ifp, m, eh);
872 		return;
873 #else
874 		goto drop;
875 #endif
876 	}
877 
878 	/* Strip off the Ethernet header. */
879 	m_adj(m, ehlen);
880 
881 	switch (etype) {
882 #ifdef INET
883 	case ETHERTYPE_IP:
884 #ifdef GATEWAY
885 		if (ipflow_fastforward(m))
886 			return;
887 #endif
888 		pktq = ip_pktq;
889 		break;
890 
891 	case ETHERTYPE_ARP:
892 		isr = NETISR_ARP;
893 		inq = &arpintrq;
894 		break;
895 
896 	case ETHERTYPE_REVARP:
897 		revarpinput(m);	/* XXX queue? */
898 		return;
899 #endif
900 
901 #ifdef INET6
902 	case ETHERTYPE_IPV6:
903 		if (__predict_false(!in6_present))
904 			goto drop;
905 #ifdef GATEWAY
906 		if (ip6flow_fastforward(&m))
907 			return;
908 #endif
909 		pktq = ip6_pktq;
910 		break;
911 #endif
912 
913 #ifdef NETATALK
914 	case ETHERTYPE_ATALK:
915 		isr = NETISR_ATALK;
916 		inq = &atintrq1;
917 		break;
918 
919 	case ETHERTYPE_AARP:
920 		aarpinput(ifp, m); /* XXX queue? */
921 		return;
922 #endif
923 
924 #ifdef MPLS
925 	case ETHERTYPE_MPLS:
926 		isr = NETISR_MPLS;
927 		inq = &mplsintrq;
928 		break;
929 #endif
930 
931 	default:
932 		goto drop;
933 	}
934 
935 	if (__predict_true(pktq)) {
936 #ifdef NET_MPSAFE
937 		const u_int h = curcpu()->ci_index;
938 #else
939 		const uint32_t h = pktq_rps_hash(m);
940 #endif
941 		if (__predict_false(!pktq_enqueue(pktq, m, h))) {
942 			m_freem(m);
943 		}
944 		return;
945 	}
946 
947 	if (__predict_false(!inq)) {
948 		/* Should not happen. */
949 		goto drop;
950 	}
951 
952 	IFQ_ENQUEUE_ISR(inq, m, isr);
953 	return;
954 
955 drop:
956 	m_freem(m);
957 dropped:
958 	if_statinc(ifp, if_ierrors); /* XXX should have a dedicated counter? */
959 }
960 
961 /*
962  * Convert Ethernet address to printable (loggable) representation.
963  */
964 char *
965 ether_sprintf(const u_char *ap)
966 {
967 	static char etherbuf[3 * ETHER_ADDR_LEN];
968 	return ether_snprintf(etherbuf, sizeof(etherbuf), ap);
969 }
970 
971 char *
972 ether_snprintf(char *buf, size_t len, const u_char *ap)
973 {
974 	char *cp = buf;
975 	size_t i;
976 
977 	for (i = 0; i < len / 3; i++) {
978 		*cp++ = hexdigits[*ap >> 4];
979 		*cp++ = hexdigits[*ap++ & 0xf];
980 		*cp++ = ':';
981 	}
982 	*--cp = '\0';
983 	return buf;
984 }
985 
986 static void
987 ether_link_state_changed(struct ifnet *ifp, int link_state)
988 {
989 #if NVLAN > 0
990 	struct ethercom *ec = (void *)ifp;
991 
992 	if (ec->ec_nvlans)
993 		vlan_link_state_changed(ifp, link_state);
994 #endif
995 }
996 
997 /*
998  * Perform common duties while attaching to interface list
999  */
1000 void
1001 ether_ifattach(struct ifnet *ifp, const uint8_t *lla)
1002 {
1003 	struct ethercom *ec = (struct ethercom *)ifp;
1004 
1005 	ifp->if_type = IFT_ETHER;
1006 	ifp->if_hdrlen = ETHER_HDR_LEN;
1007 	ifp->if_dlt = DLT_EN10MB;
1008 	ifp->if_mtu = ETHERMTU;
1009 	ifp->if_output = ether_output;
1010 	ifp->_if_input = ether_input;
1011 	ifp->if_link_state_changed = ether_link_state_changed;
1012 	if (ifp->if_baudrate == 0)
1013 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
1014 
1015 	if (lla != NULL)
1016 		if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla));
1017 
1018 	LIST_INIT(&ec->ec_multiaddrs);
1019 	SIMPLEQ_INIT(&ec->ec_vids);
1020 	ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
1021 	ec->ec_flags = 0;
1022 	ifp->if_broadcastaddr = etherbroadcastaddr;
1023 	bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header));
1024 #ifdef MBUFTRACE
1025 	mowner_init_owner(&ec->ec_tx_mowner, ifp->if_xname, "tx");
1026 	mowner_init_owner(&ec->ec_rx_mowner, ifp->if_xname, "rx");
1027 	MOWNER_ATTACH(&ec->ec_tx_mowner);
1028 	MOWNER_ATTACH(&ec->ec_rx_mowner);
1029 	ifp->if_mowner = &ec->ec_tx_mowner;
1030 #endif
1031 }
1032 
1033 void
1034 ether_ifdetach(struct ifnet *ifp)
1035 {
1036 	struct ethercom *ec = (void *) ifp;
1037 	struct ether_multi *enm;
1038 
1039 	IFNET_ASSERT_UNLOCKED(ifp);
1040 	/*
1041 	 * Prevent further calls to ioctl (for example turning off
1042 	 * promiscuous mode from the bridge code), which eventually can
1043 	 * call if_init() which can cause panics because the interface
1044 	 * is in the process of being detached. Return device not configured
1045 	 * instead.
1046 	 */
1047 	ifp->if_ioctl = __FPTRCAST(int (*)(struct ifnet *, u_long, void *),
1048 	    enxio);
1049 
1050 #if NBRIDGE > 0
1051 	if (ifp->if_bridge)
1052 		bridge_ifdetach(ifp);
1053 #endif
1054 	bpf_detach(ifp);
1055 #if NVLAN > 0
1056 	if (ec->ec_nvlans)
1057 		vlan_ifdetach(ifp);
1058 #endif
1059 
1060 	ETHER_LOCK(ec);
1061 	KASSERT(ec->ec_nvlans == 0);
1062 	while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) {
1063 		LIST_REMOVE(enm, enm_list);
1064 		kmem_free(enm, sizeof(*enm));
1065 		ec->ec_multicnt--;
1066 	}
1067 	ETHER_UNLOCK(ec);
1068 
1069 	mutex_obj_free(ec->ec_lock);
1070 	ec->ec_lock = NULL;
1071 
1072 	ifp->if_mowner = NULL;
1073 	MOWNER_DETACH(&ec->ec_rx_mowner);
1074 	MOWNER_DETACH(&ec->ec_tx_mowner);
1075 }
1076 
1077 #if 0
1078 /*
1079  * This is for reference.  We have a table-driven version
1080  * of the little-endian crc32 generator, which is faster
1081  * than the double-loop.
1082  */
1083 uint32_t
1084 ether_crc32_le(const uint8_t *buf, size_t len)
1085 {
1086 	uint32_t c, crc, carry;
1087 	size_t i, j;
1088 
1089 	crc = 0xffffffffU;	/* initial value */
1090 
1091 	for (i = 0; i < len; i++) {
1092 		c = buf[i];
1093 		for (j = 0; j < 8; j++) {
1094 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1095 			crc >>= 1;
1096 			c >>= 1;
1097 			if (carry)
1098 				crc = (crc ^ ETHER_CRC_POLY_LE);
1099 		}
1100 	}
1101 
1102 	return (crc);
1103 }
1104 #else
1105 uint32_t
1106 ether_crc32_le(const uint8_t *buf, size_t len)
1107 {
1108 	static const uint32_t crctab[] = {
1109 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1110 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1111 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1112 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1113 	};
1114 	uint32_t crc;
1115 	size_t i;
1116 
1117 	crc = 0xffffffffU;	/* initial value */
1118 
1119 	for (i = 0; i < len; i++) {
1120 		crc ^= buf[i];
1121 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1122 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1123 	}
1124 
1125 	return (crc);
1126 }
1127 #endif
1128 
1129 uint32_t
1130 ether_crc32_be(const uint8_t *buf, size_t len)
1131 {
1132 	uint32_t c, crc, carry;
1133 	size_t i, j;
1134 
1135 	crc = 0xffffffffU;	/* initial value */
1136 
1137 	for (i = 0; i < len; i++) {
1138 		c = buf[i];
1139 		for (j = 0; j < 8; j++) {
1140 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1141 			crc <<= 1;
1142 			c >>= 1;
1143 			if (carry)
1144 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1145 		}
1146 	}
1147 
1148 	return (crc);
1149 }
1150 
1151 #ifdef INET
1152 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] =
1153     { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 };
1154 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] =
1155     { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff };
1156 #endif
1157 #ifdef INET6
1158 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] =
1159     { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 };
1160 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] =
1161     { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff };
1162 #endif
1163 
1164 /*
1165  * ether_aton implementation, not using a static buffer.
1166  */
1167 int
1168 ether_aton_r(u_char *dest, size_t len, const char *str)
1169 {
1170 	const u_char *cp = (const void *)str;
1171 	u_char *ep;
1172 
1173 #define atox(c)	(((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10))
1174 
1175 	if (len < ETHER_ADDR_LEN)
1176 		return ENOSPC;
1177 
1178 	ep = dest + ETHER_ADDR_LEN;
1179 
1180 	while (*cp) {
1181 		if (!isxdigit(*cp))
1182 			return EINVAL;
1183 
1184 		*dest = atox(*cp);
1185 		cp++;
1186 		if (isxdigit(*cp)) {
1187 			*dest = (*dest << 4) | atox(*cp);
1188 			cp++;
1189 		}
1190 		dest++;
1191 
1192 		if (dest == ep)
1193 			return (*cp == '\0') ? 0 : ENAMETOOLONG;
1194 
1195 		switch (*cp) {
1196 		case ':':
1197 		case '-':
1198 		case '.':
1199 			cp++;
1200 			break;
1201 		}
1202 	}
1203 	return ENOBUFS;
1204 }
1205 
1206 /*
1207  * Convert a sockaddr into an Ethernet address or range of Ethernet
1208  * addresses.
1209  */
1210 int
1211 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN],
1212     uint8_t addrhi[ETHER_ADDR_LEN])
1213 {
1214 #ifdef INET
1215 	const struct sockaddr_in *sin;
1216 #endif
1217 #ifdef INET6
1218 	const struct sockaddr_in6 *sin6;
1219 #endif
1220 
1221 	switch (sa->sa_family) {
1222 
1223 	case AF_UNSPEC:
1224 		memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN);
1225 		memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1226 		break;
1227 
1228 #ifdef INET
1229 	case AF_INET:
1230 		sin = satocsin(sa);
1231 		if (sin->sin_addr.s_addr == INADDR_ANY) {
1232 			/*
1233 			 * An IP address of INADDR_ANY means listen to
1234 			 * or stop listening to all of the Ethernet
1235 			 * multicast addresses used for IP.
1236 			 * (This is for the sake of IP multicast routers.)
1237 			 */
1238 			memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN);
1239 			memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN);
1240 		} else {
1241 			ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo);
1242 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1243 		}
1244 		break;
1245 #endif
1246 #ifdef INET6
1247 	case AF_INET6:
1248 		sin6 = satocsin6(sa);
1249 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1250 			/*
1251 			 * An IP6 address of 0 means listen to or stop
1252 			 * listening to all of the Ethernet multicast
1253 			 * address used for IP6.
1254 			 * (This is used for multicast routers.)
1255 			 */
1256 			memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN);
1257 			memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN);
1258 		} else {
1259 			ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo);
1260 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1261 		}
1262 		break;
1263 #endif
1264 
1265 	default:
1266 		return EAFNOSUPPORT;
1267 	}
1268 	return 0;
1269 }
1270 
1271 /*
1272  * Add an Ethernet multicast address or range of addresses to the list for a
1273  * given interface.
1274  */
1275 int
1276 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec)
1277 {
1278 	struct ether_multi *enm, *_enm;
1279 	u_char addrlo[ETHER_ADDR_LEN];
1280 	u_char addrhi[ETHER_ADDR_LEN];
1281 	int error = 0;
1282 
1283 	/* Allocate out of lock */
1284 	enm = kmem_alloc(sizeof(*enm), KM_SLEEP);
1285 
1286 	ETHER_LOCK(ec);
1287 	error = ether_multiaddr(sa, addrlo, addrhi);
1288 	if (error != 0)
1289 		goto out;
1290 
1291 	/*
1292 	 * Verify that we have valid Ethernet multicast addresses.
1293 	 */
1294 	if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) {
1295 		error = EINVAL;
1296 		goto out;
1297 	}
1298 
1299 	/*
1300 	 * See if the address range is already in the list.
1301 	 */
1302 	_enm = ether_lookup_multi(addrlo, addrhi, ec);
1303 	if (_enm != NULL) {
1304 		/*
1305 		 * Found it; just increment the reference count.
1306 		 */
1307 		++_enm->enm_refcount;
1308 		error = 0;
1309 		goto out;
1310 	}
1311 
1312 	/*
1313 	 * Link a new multicast record into the interface's multicast list.
1314 	 */
1315 	memcpy(enm->enm_addrlo, addrlo, ETHER_ADDR_LEN);
1316 	memcpy(enm->enm_addrhi, addrhi, ETHER_ADDR_LEN);
1317 	enm->enm_refcount = 1;
1318 	LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list);
1319 	ec->ec_multicnt++;
1320 
1321 	/*
1322 	 * Return ENETRESET to inform the driver that the list has changed
1323 	 * and its reception filter should be adjusted accordingly.
1324 	 */
1325 	error = ENETRESET;
1326 	enm = NULL;
1327 
1328 out:
1329 	ETHER_UNLOCK(ec);
1330 	if (enm != NULL)
1331 		kmem_free(enm, sizeof(*enm));
1332 	return error;
1333 }
1334 
1335 /*
1336  * Delete a multicast address record.
1337  */
1338 int
1339 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec)
1340 {
1341 	struct ether_multi *enm;
1342 	u_char addrlo[ETHER_ADDR_LEN];
1343 	u_char addrhi[ETHER_ADDR_LEN];
1344 	int error;
1345 
1346 	ETHER_LOCK(ec);
1347 	error = ether_multiaddr(sa, addrlo, addrhi);
1348 	if (error != 0)
1349 		goto error;
1350 
1351 	/*
1352 	 * Look up the address in our list.
1353 	 */
1354 	enm = ether_lookup_multi(addrlo, addrhi, ec);
1355 	if (enm == NULL) {
1356 		error = ENXIO;
1357 		goto error;
1358 	}
1359 	if (--enm->enm_refcount != 0) {
1360 		/*
1361 		 * Still some claims to this record.
1362 		 */
1363 		error = 0;
1364 		goto error;
1365 	}
1366 
1367 	/*
1368 	 * No remaining claims to this record; unlink and free it.
1369 	 */
1370 	LIST_REMOVE(enm, enm_list);
1371 	ec->ec_multicnt--;
1372 	ETHER_UNLOCK(ec);
1373 	kmem_free(enm, sizeof(*enm));
1374 
1375 	/*
1376 	 * Return ENETRESET to inform the driver that the list has changed
1377 	 * and its reception filter should be adjusted accordingly.
1378 	 */
1379 	return ENETRESET;
1380 
1381 error:
1382 	ETHER_UNLOCK(ec);
1383 	return error;
1384 }
1385 
1386 void
1387 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb)
1388 {
1389 	ec->ec_ifflags_cb = cb;
1390 }
1391 
1392 void
1393 ether_set_vlan_cb(struct ethercom *ec, ether_vlancb_t cb)
1394 {
1395 
1396 	ec->ec_vlan_cb = cb;
1397 }
1398 
1399 static int
1400 ether_ioctl_reinit(struct ethercom *ec)
1401 {
1402 	struct ifnet *ifp = &ec->ec_if;
1403 	int error;
1404 
1405 	switch (ifp->if_flags & (IFF_UP | IFF_RUNNING)) {
1406 	case IFF_RUNNING:
1407 		/*
1408 		 * If interface is marked down and it is running,
1409 		 * then stop and disable it.
1410 		 */
1411 		(*ifp->if_stop)(ifp, 1);
1412 		break;
1413 	case IFF_UP:
1414 		/*
1415 		 * If interface is marked up and it is stopped, then
1416 		 * start it.
1417 		 */
1418 		return (*ifp->if_init)(ifp);
1419 	case IFF_UP | IFF_RUNNING:
1420 		error = 0;
1421 		if (ec->ec_ifflags_cb != NULL) {
1422 			error = (*ec->ec_ifflags_cb)(ec);
1423 			if (error == ENETRESET) {
1424 				/*
1425 				 * Reset the interface to pick up
1426 				 * changes in any other flags that
1427 				 * affect the hardware state.
1428 				 */
1429 				return (*ifp->if_init)(ifp);
1430 			}
1431 		} else
1432 			error = (*ifp->if_init)(ifp);
1433 		return error;
1434 	case 0:
1435 		break;
1436 	}
1437 
1438 	return 0;
1439 }
1440 
1441 /*
1442  * Common ioctls for Ethernet interfaces.  Note, we must be
1443  * called at splnet().
1444  */
1445 int
1446 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data)
1447 {
1448 	struct ethercom *ec = (void *)ifp;
1449 	struct eccapreq *eccr;
1450 	struct ifreq *ifr = (struct ifreq *)data;
1451 	struct if_laddrreq *iflr = data;
1452 	const struct sockaddr_dl *sdl;
1453 	static const uint8_t zero[ETHER_ADDR_LEN];
1454 	int error;
1455 
1456 	switch (cmd) {
1457 	case SIOCINITIFADDR:
1458 	    {
1459 		struct ifaddr *ifa = (struct ifaddr *)data;
1460 		if (ifa->ifa_addr->sa_family != AF_LINK
1461 		    && (ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
1462 		       (IFF_UP | IFF_RUNNING)) {
1463 			ifp->if_flags |= IFF_UP;
1464 			if ((error = (*ifp->if_init)(ifp)) != 0)
1465 				return error;
1466 		}
1467 #ifdef INET
1468 		if (ifa->ifa_addr->sa_family == AF_INET)
1469 			arp_ifinit(ifp, ifa);
1470 #endif
1471 		return 0;
1472 	    }
1473 
1474 	case SIOCSIFMTU:
1475 	    {
1476 		int maxmtu;
1477 
1478 		if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU)
1479 			maxmtu = ETHERMTU_JUMBO;
1480 		else
1481 			maxmtu = ETHERMTU;
1482 
1483 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu)
1484 			return EINVAL;
1485 		else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET)
1486 			return error;
1487 		else if (ifp->if_flags & IFF_UP) {
1488 			/* Make sure the device notices the MTU change. */
1489 			return (*ifp->if_init)(ifp);
1490 		} else
1491 			return 0;
1492 	    }
1493 
1494 	case SIOCSIFFLAGS:
1495 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
1496 			return error;
1497 		return ether_ioctl_reinit(ec);
1498 	case SIOCGIFFLAGS:
1499 		error = ifioctl_common(ifp, cmd, data);
1500 		if (error == 0) {
1501 			/* Set IFF_ALLMULTI for backcompat */
1502 			ifr->ifr_flags |= (ec->ec_flags & ETHER_F_ALLMULTI) ?
1503 			    IFF_ALLMULTI : 0;
1504 		}
1505 		return error;
1506 	case SIOCGETHERCAP:
1507 		eccr = (struct eccapreq *)data;
1508 		eccr->eccr_capabilities = ec->ec_capabilities;
1509 		eccr->eccr_capenable = ec->ec_capenable;
1510 		return 0;
1511 	case SIOCSETHERCAP:
1512 		eccr = (struct eccapreq *)data;
1513 		if ((eccr->eccr_capenable & ~ec->ec_capabilities) != 0)
1514 			return EINVAL;
1515 		if (eccr->eccr_capenable == ec->ec_capenable)
1516 			return 0;
1517 #if 0 /* notyet */
1518 		ec->ec_capenable = (ec->ec_capenable & ETHERCAP_CANTCHANGE)
1519 		    | (eccr->eccr_capenable & ~ETHERCAP_CANTCHANGE);
1520 #else
1521 		ec->ec_capenable = eccr->eccr_capenable;
1522 #endif
1523 		return ether_ioctl_reinit(ec);
1524 	case SIOCADDMULTI:
1525 		return ether_addmulti(ifreq_getaddr(cmd, ifr), ec);
1526 	case SIOCDELMULTI:
1527 		return ether_delmulti(ifreq_getaddr(cmd, ifr), ec);
1528 	case SIOCSIFMEDIA:
1529 	case SIOCGIFMEDIA:
1530 		if (ec->ec_mii != NULL)
1531 			return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media,
1532 			    cmd);
1533 		else if (ec->ec_ifmedia != NULL)
1534 			return ifmedia_ioctl(ifp, ifr, ec->ec_ifmedia, cmd);
1535 		else
1536 			return ENOTTY;
1537 		break;
1538 	case SIOCALIFADDR:
1539 		sdl = satocsdl(sstocsa(&iflr->addr));
1540 		if (sdl->sdl_family != AF_LINK)
1541 			;
1542 		else if (ETHER_IS_MULTICAST(CLLADDR(sdl)))
1543 			return EINVAL;
1544 		else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0)
1545 			return EINVAL;
1546 		/*FALLTHROUGH*/
1547 	default:
1548 		return ifioctl_common(ifp, cmd, data);
1549 	}
1550 	return 0;
1551 }
1552 
1553 /*
1554  * Enable/disable passing VLAN packets if the parent interface supports it.
1555  * Return:
1556  * 	 0: Ok
1557  *	-1: Parent interface does not support vlans
1558  *	>0: Error
1559  */
1560 int
1561 ether_enable_vlan_mtu(struct ifnet *ifp)
1562 {
1563 	int error;
1564 	struct ethercom *ec = (void *)ifp;
1565 
1566 	/* Parent does not support VLAN's */
1567 	if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0)
1568 		return -1;
1569 
1570 	/*
1571 	 * Parent supports the VLAN_MTU capability,
1572 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames;
1573 	 * enable it.
1574 	 */
1575 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1576 
1577 	/* Interface is down, defer for later */
1578 	if ((ifp->if_flags & IFF_UP) == 0)
1579 		return 0;
1580 
1581 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1582 		return 0;
1583 
1584 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1585 	return error;
1586 }
1587 
1588 int
1589 ether_disable_vlan_mtu(struct ifnet *ifp)
1590 {
1591 	int error;
1592 	struct ethercom *ec = (void *)ifp;
1593 
1594 	/* We still have VLAN's, defer for later */
1595 	if (ec->ec_nvlans != 0)
1596 		return 0;
1597 
1598 	/* Parent does not support VLAB's, nothing to do. */
1599 	if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0)
1600 		return -1;
1601 
1602 	/*
1603 	 * Disable Tx/Rx of VLAN-sized frames.
1604 	 */
1605 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1606 
1607 	/* Interface is down, defer for later */
1608 	if ((ifp->if_flags & IFF_UP) == 0)
1609 		return 0;
1610 
1611 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1612 		return 0;
1613 
1614 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1615 	return error;
1616 }
1617 
1618 static int
1619 ether_multicast_sysctl(SYSCTLFN_ARGS)
1620 {
1621 	struct ether_multi *enm;
1622 	struct ifnet *ifp;
1623 	struct ethercom *ec;
1624 	int error = 0;
1625 	size_t written;
1626 	struct psref psref;
1627 	int bound;
1628 	unsigned int multicnt;
1629 	struct ether_multi_sysctl *addrs;
1630 	int i;
1631 
1632 	if (namelen != 1)
1633 		return EINVAL;
1634 
1635 	bound = curlwp_bind();
1636 	ifp = if_get_byindex(name[0], &psref);
1637 	if (ifp == NULL) {
1638 		error = ENODEV;
1639 		goto out;
1640 	}
1641 	if (ifp->if_type != IFT_ETHER) {
1642 		if_put(ifp, &psref);
1643 		*oldlenp = 0;
1644 		goto out;
1645 	}
1646 	ec = (struct ethercom *)ifp;
1647 
1648 	if (oldp == NULL) {
1649 		if_put(ifp, &psref);
1650 		*oldlenp = ec->ec_multicnt * sizeof(*addrs);
1651 		goto out;
1652 	}
1653 
1654 	/*
1655 	 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which
1656 	 * is sleepable, while holding it. Copy data to a local buffer first
1657 	 * with the lock taken and then call sysctl_copyout without holding it.
1658 	 */
1659 retry:
1660 	multicnt = ec->ec_multicnt;
1661 
1662 	if (multicnt == 0) {
1663 		if_put(ifp, &psref);
1664 		*oldlenp = 0;
1665 		goto out;
1666 	}
1667 
1668 	addrs = kmem_zalloc(sizeof(*addrs) * multicnt, KM_SLEEP);
1669 
1670 	ETHER_LOCK(ec);
1671 	if (multicnt != ec->ec_multicnt) {
1672 		/* The number of multicast addresses has changed */
1673 		ETHER_UNLOCK(ec);
1674 		kmem_free(addrs, sizeof(*addrs) * multicnt);
1675 		goto retry;
1676 	}
1677 
1678 	i = 0;
1679 	LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) {
1680 		struct ether_multi_sysctl *addr = &addrs[i];
1681 		addr->enm_refcount = enm->enm_refcount;
1682 		memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN);
1683 		memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN);
1684 		i++;
1685 	}
1686 	ETHER_UNLOCK(ec);
1687 
1688 	error = 0;
1689 	written = 0;
1690 	for (i = 0; i < multicnt; i++) {
1691 		struct ether_multi_sysctl *addr = &addrs[i];
1692 
1693 		if (written + sizeof(*addr) > *oldlenp)
1694 			break;
1695 		error = sysctl_copyout(l, addr, oldp, sizeof(*addr));
1696 		if (error)
1697 			break;
1698 		written += sizeof(*addr);
1699 		oldp = (char *)oldp + sizeof(*addr);
1700 	}
1701 	kmem_free(addrs, sizeof(*addrs) * multicnt);
1702 
1703 	if_put(ifp, &psref);
1704 
1705 	*oldlenp = written;
1706 out:
1707 	curlwp_bindx(bound);
1708 	return error;
1709 }
1710 
1711 static void
1712 ether_sysctl_setup(struct sysctllog **clog)
1713 {
1714 	const struct sysctlnode *rnode = NULL;
1715 
1716 	sysctl_createv(clog, 0, NULL, &rnode,
1717 		       CTLFLAG_PERMANENT,
1718 		       CTLTYPE_NODE, "ether",
1719 		       SYSCTL_DESCR("Ethernet-specific information"),
1720 		       NULL, 0, NULL, 0,
1721 		       CTL_NET, CTL_CREATE, CTL_EOL);
1722 
1723 	sysctl_createv(clog, 0, &rnode, NULL,
1724 		       CTLFLAG_PERMANENT,
1725 		       CTLTYPE_NODE, "multicast",
1726 		       SYSCTL_DESCR("multicast addresses"),
1727 		       ether_multicast_sysctl, 0, NULL, 0,
1728 		       CTL_CREATE, CTL_EOL);
1729 }
1730 
1731 void
1732 etherinit(void)
1733 {
1734 
1735 #ifdef DIAGNOSTIC
1736 	mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET);
1737 #endif
1738 	ether_sysctl_setup(NULL);
1739 }
1740