xref: /netbsd-src/sys/net/if_ethersubr.c (revision 53b02e147d4ed531c0d2a5ca9b3e8026ba3e99b5)
1 /*	$NetBSD: if_ethersubr.c,v 1.307 2021/12/10 01:18:29 msaitoh 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 1.307 2021/12/10 01:18:29 msaitoh 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 #include <sys/hook.h>
93 
94 #include <net/if.h>
95 #include <net/netisr.h>
96 #include <net/route.h>
97 #include <net/if_llc.h>
98 #include <net/if_dl.h>
99 #include <net/if_types.h>
100 #include <net/pktqueue.h>
101 
102 #include <net/if_media.h>
103 #include <dev/mii/mii.h>
104 #include <dev/mii/miivar.h>
105 
106 #if NARP == 0
107 /*
108  * XXX there should really be a way to issue this warning from within config(8)
109  */
110 #error You have included NETATALK or a pseudo-device in your configuration that depends on the presence of ethernet interfaces, but have no such interfaces configured. Check if you really need pseudo-device bridge, pppoe, vlan or options NETATALK.
111 #endif
112 
113 #include <net/bpf.h>
114 
115 #include <net/if_ether.h>
116 #include <net/if_vlanvar.h>
117 
118 #if NPPPOE > 0
119 #include <net/if_pppoe.h>
120 #endif
121 
122 #if NAGR > 0
123 #include <net/ether_slowprotocols.h>
124 #include <net/agr/ieee8023ad.h>
125 #include <net/agr/if_agrvar.h>
126 #endif
127 
128 #include <net/lagg/if_laggvar.h>
129 
130 #if NBRIDGE > 0
131 #include <net/if_bridgevar.h>
132 #endif
133 
134 #include <netinet/in.h>
135 #ifdef INET
136 #include <netinet/in_var.h>
137 #endif
138 #include <netinet/if_inarp.h>
139 
140 #ifdef INET6
141 #ifndef INET
142 #include <netinet/in.h>
143 #endif
144 #include <netinet6/in6_var.h>
145 #include <netinet6/nd6.h>
146 #endif
147 
148 #include "carp.h"
149 #if NCARP > 0
150 #include <netinet/ip_carp.h>
151 #endif
152 
153 #ifdef NETATALK
154 #include <netatalk/at.h>
155 #include <netatalk/at_var.h>
156 #include <netatalk/at_extern.h>
157 
158 #define llc_snap_org_code llc_un.type_snap.org_code
159 #define llc_snap_ether_type llc_un.type_snap.ether_type
160 
161 extern u_char	at_org_code[3];
162 extern u_char	aarp_org_code[3];
163 #endif /* NETATALK */
164 
165 #ifdef MPLS
166 #include <netmpls/mpls.h>
167 #include <netmpls/mpls_var.h>
168 #endif
169 
170 CTASSERT(sizeof(struct ether_addr) == 6);
171 CTASSERT(sizeof(struct ether_header) == 14);
172 
173 #ifdef DIAGNOSTIC
174 static struct timeval bigpktppslim_last;
175 static int bigpktppslim = 2;	/* XXX */
176 static int bigpktpps_count;
177 static kmutex_t bigpktpps_lock __cacheline_aligned;
178 #endif
179 
180 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] =
181     { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
182 const uint8_t ethermulticastaddr_slowprotocols[ETHER_ADDR_LEN] =
183     { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x02 };
184 #define senderr(e) { error = (e); goto bad;}
185 
186 static pktq_rps_hash_func_t ether_pktq_rps_hash_p;
187 
188 /* if_lagg(4) support */
189 struct mbuf *(*lagg_input_ethernet_p)(struct ifnet *, struct mbuf *);
190 
191 static int ether_output(struct ifnet *, struct mbuf *,
192     const struct sockaddr *, const struct rtentry *);
193 
194 /*
195  * Ethernet output routine.
196  * Encapsulate a packet of type family for the local net.
197  * Assumes that ifp is actually pointer to ethercom structure.
198  */
199 static int
200 ether_output(struct ifnet * const ifp0, struct mbuf * const m0,
201     const struct sockaddr * const dst, const struct rtentry *rt)
202 {
203 	uint8_t esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
204 	uint16_t etype = 0;
205 	int error = 0, hdrcmplt = 0;
206 	struct mbuf *m = m0;
207 	struct mbuf *mcopy = NULL;
208 	struct ether_header *eh;
209 	struct ifnet *ifp = ifp0;
210 #ifdef INET
211 	struct arphdr *ah;
212 #endif
213 #ifdef NETATALK
214 	struct at_ifaddr *aa;
215 #endif
216 
217 #ifdef MBUFTRACE
218 	m_claimm(m, ifp->if_mowner);
219 #endif
220 
221 #if NCARP > 0
222 	if (ifp->if_type == IFT_CARP) {
223 		struct ifaddr *ifa;
224 		int s = pserialize_read_enter();
225 
226 		/* loop back if this is going to the carp interface */
227 		if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP &&
228 		    (ifa = ifa_ifwithaddr(dst)) != NULL) {
229 			if (ifa->ifa_ifp == ifp0) {
230 				pserialize_read_exit(s);
231 				return looutput(ifp0, m, dst, rt);
232 			}
233 		}
234 		pserialize_read_exit(s);
235 
236 		ifp = ifp->if_carpdev;
237 		/* ac = (struct arpcom *)ifp; */
238 
239 		if ((ifp0->if_flags & (IFF_UP | IFF_RUNNING)) !=
240 		    (IFF_UP | IFF_RUNNING))
241 			senderr(ENETDOWN);
242 	}
243 #endif
244 
245 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
246 		senderr(ENETDOWN);
247 
248 	switch (dst->sa_family) {
249 
250 #ifdef INET
251 	case AF_INET:
252 		if (m->m_flags & M_BCAST) {
253 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
254 		} else if (m->m_flags & M_MCAST) {
255 			ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst);
256 		} else {
257 			error = arpresolve(ifp0, rt, m, dst, edst, sizeof(edst));
258 			if (error)
259 				return (error == EWOULDBLOCK) ? 0 : error;
260 		}
261 		/* If broadcasting on a simplex interface, loopback a copy */
262 		if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
263 			mcopy = m_copypacket(m, M_DONTWAIT);
264 		etype = htons(ETHERTYPE_IP);
265 		break;
266 
267 	case AF_ARP:
268 		ah = mtod(m, struct arphdr *);
269 		if (m->m_flags & M_BCAST) {
270 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
271 		} else {
272 			void *tha = ar_tha(ah);
273 
274 			if (tha == NULL) {
275 				/* fake with ARPHRD_IEEE1394 */
276 				m_freem(m);
277 				return 0;
278 			}
279 			memcpy(edst, tha, sizeof(edst));
280 		}
281 
282 		ah->ar_hrd = htons(ARPHRD_ETHER);
283 
284 		switch (ntohs(ah->ar_op)) {
285 		case ARPOP_REVREQUEST:
286 		case ARPOP_REVREPLY:
287 			etype = htons(ETHERTYPE_REVARP);
288 			break;
289 
290 		case ARPOP_REQUEST:
291 		case ARPOP_REPLY:
292 		default:
293 			etype = htons(ETHERTYPE_ARP);
294 		}
295 		break;
296 #endif
297 
298 #ifdef INET6
299 	case AF_INET6:
300 		if (m->m_flags & M_BCAST) {
301 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
302 		} else if (m->m_flags & M_MCAST) {
303 			ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr,
304 			    edst);
305 		} else {
306 			error = nd6_resolve(ifp0, rt, m, dst, edst,
307 			    sizeof(edst));
308 			if (error)
309 				return (error == EWOULDBLOCK) ? 0 : error;
310 		}
311 		etype = htons(ETHERTYPE_IPV6);
312 		break;
313 #endif
314 
315 #ifdef NETATALK
316 	case AF_APPLETALK: {
317 		struct ifaddr *ifa;
318 		int s;
319 
320 		KERNEL_LOCK(1, NULL);
321 
322 		if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) {
323 			KERNEL_UNLOCK_ONE(NULL);
324 			return 0;
325 		}
326 
327 		/*
328 		 * ifaddr is the first thing in at_ifaddr
329 		 */
330 		s = pserialize_read_enter();
331 		ifa = at_ifawithnet((const struct sockaddr_at *)dst, ifp);
332 		if (ifa == NULL) {
333 			pserialize_read_exit(s);
334 			KERNEL_UNLOCK_ONE(NULL);
335 			senderr(EADDRNOTAVAIL);
336 		}
337 		aa = (struct at_ifaddr *)ifa;
338 
339 		/*
340 		 * In the phase 2 case, we need to prepend an mbuf for the
341 		 * llc header.
342 		 */
343 		if (aa->aa_flags & AFA_PHASE2) {
344 			struct llc llc;
345 
346 			M_PREPEND(m, sizeof(struct llc), M_DONTWAIT);
347 			if (m == NULL) {
348 				pserialize_read_exit(s);
349 				KERNEL_UNLOCK_ONE(NULL);
350 				senderr(ENOBUFS);
351 			}
352 
353 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
354 			llc.llc_control = LLC_UI;
355 			memcpy(llc.llc_snap_org_code, at_org_code,
356 			    sizeof(llc.llc_snap_org_code));
357 			llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK);
358 			memcpy(mtod(m, void *), &llc, sizeof(struct llc));
359 		} else {
360 			etype = htons(ETHERTYPE_ATALK);
361 		}
362 		pserialize_read_exit(s);
363 		KERNEL_UNLOCK_ONE(NULL);
364 		break;
365 	}
366 #endif /* NETATALK */
367 
368 	case pseudo_AF_HDRCMPLT:
369 		hdrcmplt = 1;
370 		memcpy(esrc,
371 		    ((const struct ether_header *)dst->sa_data)->ether_shost,
372 		    sizeof(esrc));
373 		/* FALLTHROUGH */
374 
375 	case AF_UNSPEC:
376 		memcpy(edst,
377 		    ((const struct ether_header *)dst->sa_data)->ether_dhost,
378 		    sizeof(edst));
379 		/* AF_UNSPEC doesn't swap the byte order of the ether_type. */
380 		etype = ((const struct ether_header *)dst->sa_data)->ether_type;
381 		break;
382 
383 	default:
384 		printf("%s: can't handle af%d\n", ifp->if_xname,
385 		    dst->sa_family);
386 		senderr(EAFNOSUPPORT);
387 	}
388 
389 #ifdef MPLS
390 	{
391 		struct m_tag *mtag;
392 		mtag = m_tag_find(m, PACKET_TAG_MPLS);
393 		if (mtag != NULL) {
394 			/* Having the tag itself indicates it's MPLS */
395 			etype = htons(ETHERTYPE_MPLS);
396 			m_tag_delete(m, mtag);
397 		}
398 	}
399 #endif
400 
401 	if (mcopy)
402 		(void)looutput(ifp, mcopy, dst, rt);
403 
404 	KASSERT((m->m_flags & M_PKTHDR) != 0);
405 
406 	/*
407 	 * If no ether type is set, this must be a 802.2 formatted packet.
408 	 */
409 	if (etype == 0)
410 		etype = htons(m->m_pkthdr.len);
411 
412 	/*
413 	 * Add local net header. If no space in first mbuf, allocate another.
414 	 */
415 	M_PREPEND(m, sizeof(struct ether_header), M_DONTWAIT);
416 	if (m == NULL)
417 		senderr(ENOBUFS);
418 
419 	eh = mtod(m, struct ether_header *);
420 	/* Note: etype is already in network byte order. */
421 	memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
422 	memcpy(eh->ether_dhost, edst, sizeof(edst));
423 	if (hdrcmplt) {
424 		memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost));
425 	} else {
426 	 	memcpy(eh->ether_shost, CLLADDR(ifp->if_sadl),
427 		    sizeof(eh->ether_shost));
428 	}
429 
430 #if NCARP > 0
431 	if (ifp0 != ifp && ifp0->if_type == IFT_CARP) {
432 	 	memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl),
433 		    sizeof(eh->ether_shost));
434 	}
435 #endif
436 
437 	if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0)
438 		return error;
439 	if (m == NULL)
440 		return 0;
441 
442 #if NBRIDGE > 0
443 	/*
444 	 * Bridges require special output handling.
445 	 */
446 	if (ifp->if_bridge)
447 		return bridge_output(ifp, m, NULL, NULL);
448 #endif
449 
450 #if NCARP > 0
451 	if (ifp != ifp0)
452 		if_statadd(ifp0, if_obytes, m->m_pkthdr.len + ETHER_HDR_LEN);
453 #endif
454 
455 #ifdef ALTQ
456 	KERNEL_LOCK(1, NULL);
457 	/*
458 	 * If ALTQ is enabled on the parent interface, do
459 	 * classification; the queueing discipline might not
460 	 * require classification, but might require the
461 	 * address family/header pointer in the pktattr.
462 	 */
463 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
464 		altq_etherclassify(&ifp->if_snd, m);
465 	KERNEL_UNLOCK_ONE(NULL);
466 #endif
467 	return ifq_enqueue(ifp, m);
468 
469 bad:
470 	if_statinc(ifp, if_oerrors);
471 	if (m)
472 		m_freem(m);
473 	return error;
474 }
475 
476 #ifdef ALTQ
477 /*
478  * This routine is a slight hack to allow a packet to be classified
479  * if the Ethernet headers are present.  It will go away when ALTQ's
480  * classification engine understands link headers.
481  *
482  * XXX: We may need to do m_pullups here. First to ensure struct ether_header
483  * is indeed contiguous, then to read the LLC and so on.
484  */
485 void
486 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m)
487 {
488 	struct ether_header *eh;
489 	struct mbuf *mtop = m;
490 	uint16_t ether_type;
491 	int hlen, af, hdrsize;
492 	void *hdr;
493 
494 	KASSERT((mtop->m_flags & M_PKTHDR) != 0);
495 
496 	hlen = ETHER_HDR_LEN;
497 	eh = mtod(m, struct ether_header *);
498 
499 	ether_type = htons(eh->ether_type);
500 
501 	if (ether_type < ETHERMTU) {
502 		/* LLC/SNAP */
503 		struct llc *llc = (struct llc *)(eh + 1);
504 		hlen += 8;
505 
506 		if (m->m_len < hlen ||
507 		    llc->llc_dsap != LLC_SNAP_LSAP ||
508 		    llc->llc_ssap != LLC_SNAP_LSAP ||
509 		    llc->llc_control != LLC_UI) {
510 			/* Not SNAP. */
511 			goto bad;
512 		}
513 
514 		ether_type = htons(llc->llc_un.type_snap.ether_type);
515 	}
516 
517 	switch (ether_type) {
518 	case ETHERTYPE_IP:
519 		af = AF_INET;
520 		hdrsize = 20;		/* sizeof(struct ip) */
521 		break;
522 
523 	case ETHERTYPE_IPV6:
524 		af = AF_INET6;
525 		hdrsize = 40;		/* sizeof(struct ip6_hdr) */
526 		break;
527 
528 	default:
529 		af = AF_UNSPEC;
530 		hdrsize = 0;
531 		break;
532 	}
533 
534 	while (m->m_len <= hlen) {
535 		hlen -= m->m_len;
536 		m = m->m_next;
537 		if (m == NULL)
538 			goto bad;
539 	}
540 
541 	if (m->m_len < (hlen + hdrsize)) {
542 		/*
543 		 * protocol header not in a single mbuf.
544 		 * We can't cope with this situation right
545 		 * now (but it shouldn't ever happen, really, anyhow).
546 		 */
547 #ifdef DEBUG
548 		printf("altq_etherclassify: headers span multiple mbufs: "
549 		    "%d < %d\n", m->m_len, (hlen + hdrsize));
550 #endif
551 		goto bad;
552 	}
553 
554 	m->m_data += hlen;
555 	m->m_len -= hlen;
556 
557 	hdr = mtod(m, void *);
558 
559 	if (ALTQ_NEEDS_CLASSIFY(ifq)) {
560 		mtop->m_pkthdr.pattr_class =
561 		    (*ifq->altq_classify)(ifq->altq_clfier, m, af);
562 	}
563 	mtop->m_pkthdr.pattr_af = af;
564 	mtop->m_pkthdr.pattr_hdr = hdr;
565 
566 	m->m_data -= hlen;
567 	m->m_len += hlen;
568 
569 	return;
570 
571 bad:
572 	mtop->m_pkthdr.pattr_class = NULL;
573 	mtop->m_pkthdr.pattr_hdr = NULL;
574 	mtop->m_pkthdr.pattr_af = AF_UNSPEC;
575 }
576 #endif /* ALTQ */
577 
578 #if defined (LLC) || defined (NETATALK)
579 static void
580 ether_input_llc(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh)
581 {
582 	struct ifqueue *inq = NULL;
583 	int isr = 0;
584 	struct llc *l;
585 
586 	if (m->m_len < sizeof(*eh) + sizeof(struct llc))
587 		goto error;
588 
589 	l = (struct llc *)(eh+1);
590 	switch (l->llc_dsap) {
591 #ifdef NETATALK
592 	case LLC_SNAP_LSAP:
593 		switch (l->llc_control) {
594 		case LLC_UI:
595 			if (l->llc_ssap != LLC_SNAP_LSAP)
596 				goto error;
597 
598 			if (memcmp(&(l->llc_snap_org_code)[0],
599 			    at_org_code, sizeof(at_org_code)) == 0 &&
600 			    ntohs(l->llc_snap_ether_type) ==
601 			    ETHERTYPE_ATALK) {
602 				inq = &atintrq2;
603 				m_adj(m, sizeof(struct ether_header)
604 				    + sizeof(struct llc));
605 				isr = NETISR_ATALK;
606 				break;
607 			}
608 
609 			if (memcmp(&(l->llc_snap_org_code)[0],
610 			    aarp_org_code,
611 			    sizeof(aarp_org_code)) == 0 &&
612 			    ntohs(l->llc_snap_ether_type) ==
613 			    ETHERTYPE_AARP) {
614 				m_adj(m, sizeof(struct ether_header)
615 				    + sizeof(struct llc));
616 				aarpinput(ifp, m); /* XXX queue? */
617 				return;
618 			}
619 
620 		default:
621 			goto error;
622 		}
623 		break;
624 #endif
625 	default:
626 		goto noproto;
627 	}
628 
629 	KASSERT(inq != NULL);
630 	IFQ_ENQUEUE_ISR(inq, m, isr);
631 	return;
632 
633 noproto:
634 	m_freem(m);
635 	if_statinc(ifp, if_noproto);
636 	return;
637 error:
638 	m_freem(m);
639 	if_statinc(ifp, if_ierrors);
640 	return;
641 }
642 #endif /* defined (LLC) || defined (NETATALK) */
643 
644 /*
645  * Process a received Ethernet packet;
646  * the packet is in the mbuf chain m with
647  * the ether header.
648  */
649 void
650 ether_input(struct ifnet *ifp, struct mbuf *m)
651 {
652 	struct ethercom *ec = (struct ethercom *) ifp;
653 	pktqueue_t *pktq = NULL;
654 	struct ifqueue *inq = NULL;
655 	uint16_t etype;
656 	struct ether_header *eh;
657 	size_t ehlen;
658 	static int earlypkts;
659 	int isr = 0;
660 #if NAGR > 0
661 	void *agrprivate;
662 #endif
663 
664 	KASSERT(!cpu_intr_p());
665 	KASSERT((m->m_flags & M_PKTHDR) != 0);
666 
667 	if ((ifp->if_flags & IFF_UP) == 0)
668 		goto drop;
669 
670 #ifdef MBUFTRACE
671 	m_claimm(m, &ec->ec_rx_mowner);
672 #endif
673 
674 	if (__predict_false(m->m_len < sizeof(*eh))) {
675 		if ((m = m_pullup(m, sizeof(*eh))) == NULL) {
676 			if_statinc(ifp, if_ierrors);
677 			return;
678 		}
679 	}
680 
681 	eh = mtod(m, struct ether_header *);
682 	etype = ntohs(eh->ether_type);
683 	ehlen = sizeof(*eh);
684 
685 	if (__predict_false(earlypkts < 100 ||
686 		entropy_epoch() == (unsigned)-1)) {
687 		rnd_add_data(NULL, eh, ehlen, 0);
688 		earlypkts++;
689 	}
690 
691 	/*
692 	 * Determine if the packet is within its size limits. For MPLS the
693 	 * header length is variable, so we skip the check.
694 	 */
695 	if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len >
696 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
697 #ifdef DIAGNOSTIC
698 		mutex_enter(&bigpktpps_lock);
699 		if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count,
700 		    bigpktppslim)) {
701 			printf("%s: discarding oversize frame (len=%d)\n",
702 			    ifp->if_xname, m->m_pkthdr.len);
703 		}
704 		mutex_exit(&bigpktpps_lock);
705 #endif
706 		goto error;
707 	}
708 
709 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
710 		/*
711 		 * If this is not a simplex interface, drop the packet
712 		 * if it came from us.
713 		 */
714 		if ((ifp->if_flags & IFF_SIMPLEX) == 0 &&
715 		    memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost,
716 		    ETHER_ADDR_LEN) == 0) {
717 			goto drop;
718 		}
719 
720 		if (memcmp(etherbroadcastaddr,
721 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
722 			m->m_flags |= M_BCAST;
723 		else
724 			m->m_flags |= M_MCAST;
725 		if_statinc(ifp, if_imcasts);
726 	}
727 
728 	/* If the CRC is still on the packet, trim it off. */
729 	if (m->m_flags & M_HASFCS) {
730 		m_adj(m, -ETHER_CRC_LEN);
731 		m->m_flags &= ~M_HASFCS;
732 	}
733 
734 	if_statadd(ifp, if_ibytes, m->m_pkthdr.len);
735 
736 #if NCARP > 0
737 	if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) {
738 		/*
739 		 * Clear M_PROMISC, in case the packet comes from a
740 		 * vlan.
741 		 */
742 		m->m_flags &= ~M_PROMISC;
743 		if (carp_input(m, (uint8_t *)&eh->ether_shost,
744 		    (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0)
745 			return;
746 	}
747 #endif
748 
749 	if ((m->m_flags & (M_BCAST | M_MCAST | M_PROMISC)) == 0 &&
750 	    (ifp->if_flags & IFF_PROMISC) != 0 &&
751 	    memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost,
752 	     ETHER_ADDR_LEN) != 0) {
753 		m->m_flags |= M_PROMISC;
754 	}
755 
756 	if ((m->m_flags & M_PROMISC) == 0) {
757 		if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
758 			return;
759 		if (m == NULL)
760 			return;
761 
762 		eh = mtod(m, struct ether_header *);
763 		etype = ntohs(eh->ether_type);
764 	}
765 
766 #if NAGR > 0
767 	if (ifp->if_type != IFT_IEEE8023ADLAG) {
768 		agrprivate = ifp->if_lagg;
769 	} else {
770 		agrprivate = NULL;
771 	}
772 	if (agrprivate != NULL &&
773 	    __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) {
774 		m->m_flags &= ~M_PROMISC;
775 		agr_input(ifp, m);
776 		return;
777 	}
778 #endif
779 
780 	/* Handle input from a lagg(4) port */
781 	if (ifp->if_type == IFT_IEEE8023ADLAG) {
782 		KASSERT(lagg_input_ethernet_p != NULL);
783 		m = (*lagg_input_ethernet_p)(ifp, m);
784 		if (m == NULL)
785 			return;
786 	}
787 
788 	/*
789 	 * If VLANs are configured on the interface, check to
790 	 * see if the device performed the decapsulation and
791 	 * provided us with the tag.
792 	 */
793 	if (ec->ec_nvlans && vlan_has_tag(m)) {
794 #if NVLAN > 0
795 		/*
796 		 * vlan_input() will either recursively call ether_input()
797 		 * or drop the packet.
798 		 */
799 		vlan_input(ifp, m);
800 		return;
801 #else
802 		goto noproto;
803 #endif
804 	}
805 
806 	/*
807 	 * Handle protocols that expect to have the Ethernet header
808 	 * (and possibly FCS) intact.
809 	 */
810 	switch (etype) {
811 	case ETHERTYPE_VLAN: {
812 		struct ether_vlan_header *evl = (void *)eh;
813 
814 		/*
815 		 * If there is a tag of 0, then the VLAN header was probably
816 		 * just being used to store the priority.  Extract the ether
817 		 * type, and if IP or IPV6, let them deal with it.
818 		 */
819 		if (m->m_len >= sizeof(*evl) &&
820 		    EVL_VLANOFTAG(ntohs(evl->evl_tag)) == 0) {
821 			etype = ntohs(evl->evl_proto);
822 			ehlen = sizeof(*evl);
823 			if ((m->m_flags & M_PROMISC) == 0 &&
824 			    (etype == ETHERTYPE_IP ||
825 			     etype == ETHERTYPE_IPV6))
826 				break;
827 		}
828 
829 #if NVLAN > 0
830 		/*
831 		 * vlan_input() will either recursively call ether_input()
832 		 * or drop the packet.
833 		 */
834 		if (ec->ec_nvlans != 0) {
835 			vlan_input(ifp, m);
836 			return;
837 		} else
838 #endif
839 			goto noproto;
840 	}
841 
842 #if NPPPOE > 0
843 	case ETHERTYPE_PPPOEDISC:
844 		pppoedisc_input(ifp, m);
845 		return;
846 
847 	case ETHERTYPE_PPPOE:
848 		pppoe_input(ifp, m);
849 		return;
850 #endif
851 
852 	case ETHERTYPE_SLOWPROTOCOLS: {
853 		uint8_t subtype;
854 
855 		if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype))
856 			goto error;
857 
858 		m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype);
859 		switch (subtype) {
860 #if NAGR > 0
861 		case SLOWPROTOCOLS_SUBTYPE_LACP:
862 			if (agrprivate != NULL) {
863 				ieee8023ad_lacp_input(ifp, m);
864 				return;
865 			}
866 			break;
867 
868 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
869 			if (agrprivate != NULL) {
870 				ieee8023ad_marker_input(ifp, m);
871 				return;
872 			}
873 			break;
874 #endif
875 
876 		default:
877 			if (subtype == 0 || subtype > 10) {
878 				/* illegal value */
879 				goto error;
880 			}
881 			/* unknown subtype */
882 			break;
883 		}
884 	}
885 	/* FALLTHROUGH */
886 	default:
887 		if (m->m_flags & M_PROMISC)
888 			goto drop;
889 	}
890 
891 	/* If the CRC is still on the packet, trim it off. */
892 	if (m->m_flags & M_HASFCS) {
893 		m_adj(m, -ETHER_CRC_LEN);
894 		m->m_flags &= ~M_HASFCS;
895 	}
896 
897 	/* etype represents the size of the payload in this case */
898 	if (etype <= ETHERMTU + sizeof(struct ether_header)) {
899 		KASSERT(ehlen == sizeof(*eh));
900 #if defined (LLC) || defined (NETATALK)
901 		ether_input_llc(ifp, m, eh);
902 		return;
903 #else
904 		/* ethertype of 0-1500 is regarded as noproto */
905 		goto noproto;
906 #endif
907 	}
908 
909 	/* Strip off the Ethernet header. */
910 	m_adj(m, ehlen);
911 
912 	switch (etype) {
913 #ifdef INET
914 	case ETHERTYPE_IP:
915 #ifdef GATEWAY
916 		if (ipflow_fastforward(m))
917 			return;
918 #endif
919 		pktq = ip_pktq;
920 		break;
921 
922 	case ETHERTYPE_ARP:
923 		isr = NETISR_ARP;
924 		inq = &arpintrq;
925 		break;
926 
927 	case ETHERTYPE_REVARP:
928 		revarpinput(m);	/* XXX queue? */
929 		return;
930 #endif
931 
932 #ifdef INET6
933 	case ETHERTYPE_IPV6:
934 		if (__predict_false(!in6_present))
935 			goto noproto;
936 #ifdef GATEWAY
937 		if (ip6flow_fastforward(&m))
938 			return;
939 #endif
940 		pktq = ip6_pktq;
941 		break;
942 #endif
943 
944 #ifdef NETATALK
945 	case ETHERTYPE_ATALK:
946 		isr = NETISR_ATALK;
947 		inq = &atintrq1;
948 		break;
949 
950 	case ETHERTYPE_AARP:
951 		aarpinput(ifp, m); /* XXX queue? */
952 		return;
953 #endif
954 
955 #ifdef MPLS
956 	case ETHERTYPE_MPLS:
957 		isr = NETISR_MPLS;
958 		inq = &mplsintrq;
959 		break;
960 #endif
961 
962 	default:
963 		goto noproto;
964 	}
965 
966 	if (__predict_true(pktq)) {
967 		const uint32_t h = pktq_rps_hash(&ether_pktq_rps_hash_p, m);
968 		if (__predict_false(!pktq_enqueue(pktq, m, h))) {
969 			m_freem(m);
970 		}
971 		return;
972 	}
973 
974 	if (__predict_false(!inq)) {
975 		/* Should not happen. */
976 		goto error;
977 	}
978 
979 	IFQ_ENQUEUE_ISR(inq, m, isr);
980 	return;
981 
982 drop:
983 	m_freem(m);
984 	if_statinc(ifp, if_iqdrops);
985 	return;
986 noproto:
987 	m_freem(m);
988 	if_statinc(ifp, if_noproto);
989 	return;
990 error:
991 	m_freem(m);
992 	if_statinc(ifp, if_ierrors);
993 	return;
994 }
995 
996 /*
997  * Convert Ethernet address to printable (loggable) representation.
998  */
999 char *
1000 ether_sprintf(const u_char *ap)
1001 {
1002 	static char etherbuf[3 * ETHER_ADDR_LEN];
1003 	return ether_snprintf(etherbuf, sizeof(etherbuf), ap);
1004 }
1005 
1006 char *
1007 ether_snprintf(char *buf, size_t len, const u_char *ap)
1008 {
1009 	char *cp = buf;
1010 	size_t i;
1011 
1012 	for (i = 0; i < len / 3; i++) {
1013 		*cp++ = hexdigits[*ap >> 4];
1014 		*cp++ = hexdigits[*ap++ & 0xf];
1015 		*cp++ = ':';
1016 	}
1017 	*--cp = '\0';
1018 	return buf;
1019 }
1020 
1021 /*
1022  * Perform common duties while attaching to interface list
1023  */
1024 void
1025 ether_ifattach(struct ifnet *ifp, const uint8_t *lla)
1026 {
1027 	struct ethercom *ec = (struct ethercom *)ifp;
1028 	char xnamebuf[HOOKNAMSIZ];
1029 
1030 	ifp->if_type = IFT_ETHER;
1031 	ifp->if_hdrlen = ETHER_HDR_LEN;
1032 	ifp->if_dlt = DLT_EN10MB;
1033 	ifp->if_mtu = ETHERMTU;
1034 	ifp->if_output = ether_output;
1035 	ifp->_if_input = ether_input;
1036 	if (ifp->if_baudrate == 0)
1037 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
1038 
1039 	if (lla != NULL)
1040 		if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla));
1041 
1042 	LIST_INIT(&ec->ec_multiaddrs);
1043 	SIMPLEQ_INIT(&ec->ec_vids);
1044 	ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
1045 	ec->ec_flags = 0;
1046 	ifp->if_broadcastaddr = etherbroadcastaddr;
1047 	bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header));
1048 	snprintf(xnamebuf, sizeof(xnamebuf),
1049 	    "%s-ether_ifdetachhooks", ifp->if_xname);
1050 	ec->ec_ifdetach_hooks = simplehook_create(IPL_NET, xnamebuf);
1051 #ifdef MBUFTRACE
1052 	mowner_init_owner(&ec->ec_tx_mowner, ifp->if_xname, "tx");
1053 	mowner_init_owner(&ec->ec_rx_mowner, ifp->if_xname, "rx");
1054 	MOWNER_ATTACH(&ec->ec_tx_mowner);
1055 	MOWNER_ATTACH(&ec->ec_rx_mowner);
1056 	ifp->if_mowner = &ec->ec_tx_mowner;
1057 #endif
1058 }
1059 
1060 void
1061 ether_ifdetach(struct ifnet *ifp)
1062 {
1063 	struct ethercom *ec = (void *) ifp;
1064 	struct ether_multi *enm;
1065 
1066 	IFNET_ASSERT_UNLOCKED(ifp);
1067 	/*
1068 	 * Prevent further calls to ioctl (for example turning off
1069 	 * promiscuous mode from the bridge code), which eventually can
1070 	 * call if_init() which can cause panics because the interface
1071 	 * is in the process of being detached. Return device not configured
1072 	 * instead.
1073 	 */
1074 	ifp->if_ioctl = __FPTRCAST(int (*)(struct ifnet *, u_long, void *),
1075 	    enxio);
1076 
1077 	simplehook_dohooks(ec->ec_ifdetach_hooks);
1078 	KASSERT(!simplehook_has_hooks(ec->ec_ifdetach_hooks));
1079 	simplehook_destroy(ec->ec_ifdetach_hooks);
1080 
1081 	bpf_detach(ifp);
1082 
1083 	ETHER_LOCK(ec);
1084 	KASSERT(ec->ec_nvlans == 0);
1085 	while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) {
1086 		LIST_REMOVE(enm, enm_list);
1087 		kmem_free(enm, sizeof(*enm));
1088 		ec->ec_multicnt--;
1089 	}
1090 	ETHER_UNLOCK(ec);
1091 
1092 	mutex_obj_free(ec->ec_lock);
1093 	ec->ec_lock = NULL;
1094 
1095 	ifp->if_mowner = NULL;
1096 	MOWNER_DETACH(&ec->ec_rx_mowner);
1097 	MOWNER_DETACH(&ec->ec_tx_mowner);
1098 }
1099 
1100 void *
1101 ether_ifdetachhook_establish(struct ifnet *ifp,
1102     void (*fn)(void *), void *arg)
1103 {
1104 	struct ethercom *ec;
1105 	khook_t *hk;
1106 
1107 	if (ifp->if_type != IFT_ETHER)
1108 		return NULL;
1109 
1110 	ec = (struct ethercom *)ifp;
1111 	hk = simplehook_establish(ec->ec_ifdetach_hooks,
1112 	    fn, arg);
1113 
1114 	return (void *)hk;
1115 }
1116 
1117 void
1118 ether_ifdetachhook_disestablish(struct ifnet *ifp,
1119     void *vhook, kmutex_t *lock)
1120 {
1121 	struct ethercom *ec;
1122 
1123 	if (vhook == NULL)
1124 		return;
1125 
1126 	ec = (struct ethercom *)ifp;
1127 	simplehook_disestablish(ec->ec_ifdetach_hooks, vhook, lock);
1128 }
1129 
1130 #if 0
1131 /*
1132  * This is for reference.  We have a table-driven version
1133  * of the little-endian crc32 generator, which is faster
1134  * than the double-loop.
1135  */
1136 uint32_t
1137 ether_crc32_le(const uint8_t *buf, size_t len)
1138 {
1139 	uint32_t c, crc, carry;
1140 	size_t i, j;
1141 
1142 	crc = 0xffffffffU;	/* initial value */
1143 
1144 	for (i = 0; i < len; i++) {
1145 		c = buf[i];
1146 		for (j = 0; j < 8; j++) {
1147 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1148 			crc >>= 1;
1149 			c >>= 1;
1150 			if (carry)
1151 				crc = (crc ^ ETHER_CRC_POLY_LE);
1152 		}
1153 	}
1154 
1155 	return (crc);
1156 }
1157 #else
1158 uint32_t
1159 ether_crc32_le(const uint8_t *buf, size_t len)
1160 {
1161 	static const uint32_t crctab[] = {
1162 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1163 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1164 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1165 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1166 	};
1167 	uint32_t crc;
1168 	size_t i;
1169 
1170 	crc = 0xffffffffU;	/* initial value */
1171 
1172 	for (i = 0; i < len; i++) {
1173 		crc ^= buf[i];
1174 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1175 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1176 	}
1177 
1178 	return (crc);
1179 }
1180 #endif
1181 
1182 uint32_t
1183 ether_crc32_be(const uint8_t *buf, size_t len)
1184 {
1185 	uint32_t c, crc, carry;
1186 	size_t i, j;
1187 
1188 	crc = 0xffffffffU;	/* initial value */
1189 
1190 	for (i = 0; i < len; i++) {
1191 		c = buf[i];
1192 		for (j = 0; j < 8; j++) {
1193 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1194 			crc <<= 1;
1195 			c >>= 1;
1196 			if (carry)
1197 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1198 		}
1199 	}
1200 
1201 	return (crc);
1202 }
1203 
1204 #ifdef INET
1205 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] =
1206     { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 };
1207 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] =
1208     { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff };
1209 #endif
1210 #ifdef INET6
1211 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] =
1212     { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 };
1213 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] =
1214     { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff };
1215 #endif
1216 
1217 /*
1218  * ether_aton implementation, not using a static buffer.
1219  */
1220 int
1221 ether_aton_r(u_char *dest, size_t len, const char *str)
1222 {
1223 	const u_char *cp = (const void *)str;
1224 	u_char *ep;
1225 
1226 #define atox(c)	(((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10))
1227 
1228 	if (len < ETHER_ADDR_LEN)
1229 		return ENOSPC;
1230 
1231 	ep = dest + ETHER_ADDR_LEN;
1232 
1233 	while (*cp) {
1234 		if (!isxdigit(*cp))
1235 			return EINVAL;
1236 
1237 		*dest = atox(*cp);
1238 		cp++;
1239 		if (isxdigit(*cp)) {
1240 			*dest = (*dest << 4) | atox(*cp);
1241 			cp++;
1242 		}
1243 		dest++;
1244 
1245 		if (dest == ep)
1246 			return (*cp == '\0') ? 0 : ENAMETOOLONG;
1247 
1248 		switch (*cp) {
1249 		case ':':
1250 		case '-':
1251 		case '.':
1252 			cp++;
1253 			break;
1254 		}
1255 	}
1256 	return ENOBUFS;
1257 }
1258 
1259 /*
1260  * Convert a sockaddr into an Ethernet address or range of Ethernet
1261  * addresses.
1262  */
1263 int
1264 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN],
1265     uint8_t addrhi[ETHER_ADDR_LEN])
1266 {
1267 #ifdef INET
1268 	const struct sockaddr_in *sin;
1269 #endif
1270 #ifdef INET6
1271 	const struct sockaddr_in6 *sin6;
1272 #endif
1273 
1274 	switch (sa->sa_family) {
1275 
1276 	case AF_UNSPEC:
1277 		memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN);
1278 		memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1279 		break;
1280 
1281 #ifdef INET
1282 	case AF_INET:
1283 		sin = satocsin(sa);
1284 		if (sin->sin_addr.s_addr == INADDR_ANY) {
1285 			/*
1286 			 * An IP address of INADDR_ANY means listen to
1287 			 * or stop listening to all of the Ethernet
1288 			 * multicast addresses used for IP.
1289 			 * (This is for the sake of IP multicast routers.)
1290 			 */
1291 			memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN);
1292 			memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN);
1293 		} else {
1294 			ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo);
1295 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1296 		}
1297 		break;
1298 #endif
1299 #ifdef INET6
1300 	case AF_INET6:
1301 		sin6 = satocsin6(sa);
1302 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1303 			/*
1304 			 * An IP6 address of 0 means listen to or stop
1305 			 * listening to all of the Ethernet multicast
1306 			 * address used for IP6.
1307 			 * (This is used for multicast routers.)
1308 			 */
1309 			memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN);
1310 			memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN);
1311 		} else {
1312 			ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo);
1313 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1314 		}
1315 		break;
1316 #endif
1317 
1318 	default:
1319 		return EAFNOSUPPORT;
1320 	}
1321 	return 0;
1322 }
1323 
1324 /*
1325  * Add an Ethernet multicast address or range of addresses to the list for a
1326  * given interface.
1327  */
1328 int
1329 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec)
1330 {
1331 	struct ether_multi *enm, *_enm;
1332 	u_char addrlo[ETHER_ADDR_LEN];
1333 	u_char addrhi[ETHER_ADDR_LEN];
1334 	int error = 0;
1335 
1336 	/* Allocate out of lock */
1337 	enm = kmem_alloc(sizeof(*enm), KM_SLEEP);
1338 
1339 	ETHER_LOCK(ec);
1340 	error = ether_multiaddr(sa, addrlo, addrhi);
1341 	if (error != 0)
1342 		goto out;
1343 
1344 	/*
1345 	 * Verify that we have valid Ethernet multicast addresses.
1346 	 */
1347 	if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) {
1348 		error = EINVAL;
1349 		goto out;
1350 	}
1351 
1352 	/*
1353 	 * See if the address range is already in the list.
1354 	 */
1355 	_enm = ether_lookup_multi(addrlo, addrhi, ec);
1356 	if (_enm != NULL) {
1357 		/*
1358 		 * Found it; just increment the reference count.
1359 		 */
1360 		++_enm->enm_refcount;
1361 		error = 0;
1362 		goto out;
1363 	}
1364 
1365 	/*
1366 	 * Link a new multicast record into the interface's multicast list.
1367 	 */
1368 	memcpy(enm->enm_addrlo, addrlo, ETHER_ADDR_LEN);
1369 	memcpy(enm->enm_addrhi, addrhi, ETHER_ADDR_LEN);
1370 	enm->enm_refcount = 1;
1371 	LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list);
1372 	ec->ec_multicnt++;
1373 
1374 	/*
1375 	 * Return ENETRESET to inform the driver that the list has changed
1376 	 * and its reception filter should be adjusted accordingly.
1377 	 */
1378 	error = ENETRESET;
1379 	enm = NULL;
1380 
1381 out:
1382 	ETHER_UNLOCK(ec);
1383 	if (enm != NULL)
1384 		kmem_free(enm, sizeof(*enm));
1385 	return error;
1386 }
1387 
1388 /*
1389  * Delete a multicast address record.
1390  */
1391 int
1392 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec)
1393 {
1394 	struct ether_multi *enm;
1395 	u_char addrlo[ETHER_ADDR_LEN];
1396 	u_char addrhi[ETHER_ADDR_LEN];
1397 	int error;
1398 
1399 	ETHER_LOCK(ec);
1400 	error = ether_multiaddr(sa, addrlo, addrhi);
1401 	if (error != 0)
1402 		goto error;
1403 
1404 	/*
1405 	 * Look up the address in our list.
1406 	 */
1407 	enm = ether_lookup_multi(addrlo, addrhi, ec);
1408 	if (enm == NULL) {
1409 		error = ENXIO;
1410 		goto error;
1411 	}
1412 	if (--enm->enm_refcount != 0) {
1413 		/*
1414 		 * Still some claims to this record.
1415 		 */
1416 		error = 0;
1417 		goto error;
1418 	}
1419 
1420 	/*
1421 	 * No remaining claims to this record; unlink and free it.
1422 	 */
1423 	LIST_REMOVE(enm, enm_list);
1424 	ec->ec_multicnt--;
1425 	ETHER_UNLOCK(ec);
1426 	kmem_free(enm, sizeof(*enm));
1427 
1428 	/*
1429 	 * Return ENETRESET to inform the driver that the list has changed
1430 	 * and its reception filter should be adjusted accordingly.
1431 	 */
1432 	return ENETRESET;
1433 
1434 error:
1435 	ETHER_UNLOCK(ec);
1436 	return error;
1437 }
1438 
1439 void
1440 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb)
1441 {
1442 	ec->ec_ifflags_cb = cb;
1443 }
1444 
1445 void
1446 ether_set_vlan_cb(struct ethercom *ec, ether_vlancb_t cb)
1447 {
1448 
1449 	ec->ec_vlan_cb = cb;
1450 }
1451 
1452 static int
1453 ether_ioctl_reinit(struct ethercom *ec)
1454 {
1455 	struct ifnet *ifp = &ec->ec_if;
1456 	int error;
1457 
1458 	switch (ifp->if_flags & (IFF_UP | IFF_RUNNING)) {
1459 	case IFF_RUNNING:
1460 		/*
1461 		 * If interface is marked down and it is running,
1462 		 * then stop and disable it.
1463 		 */
1464 		(*ifp->if_stop)(ifp, 1);
1465 		break;
1466 	case IFF_UP:
1467 		/*
1468 		 * If interface is marked up and it is stopped, then
1469 		 * start it.
1470 		 */
1471 		return (*ifp->if_init)(ifp);
1472 	case IFF_UP | IFF_RUNNING:
1473 		error = 0;
1474 		if (ec->ec_ifflags_cb != NULL) {
1475 			error = (*ec->ec_ifflags_cb)(ec);
1476 			if (error == ENETRESET) {
1477 				/*
1478 				 * Reset the interface to pick up
1479 				 * changes in any other flags that
1480 				 * affect the hardware state.
1481 				 */
1482 				return (*ifp->if_init)(ifp);
1483 			}
1484 		} else
1485 			error = (*ifp->if_init)(ifp);
1486 		return error;
1487 	case 0:
1488 		break;
1489 	}
1490 
1491 	return 0;
1492 }
1493 
1494 /*
1495  * Common ioctls for Ethernet interfaces.  Note, we must be
1496  * called at splnet().
1497  */
1498 int
1499 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data)
1500 {
1501 	struct ethercom *ec = (void *)ifp;
1502 	struct eccapreq *eccr;
1503 	struct ifreq *ifr = (struct ifreq *)data;
1504 	struct if_laddrreq *iflr = data;
1505 	const struct sockaddr_dl *sdl;
1506 	static const uint8_t zero[ETHER_ADDR_LEN];
1507 	int error;
1508 
1509 	switch (cmd) {
1510 	case SIOCINITIFADDR:
1511 	    {
1512 		struct ifaddr *ifa = (struct ifaddr *)data;
1513 		if (ifa->ifa_addr->sa_family != AF_LINK
1514 		    && (ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
1515 		       (IFF_UP | IFF_RUNNING)) {
1516 			ifp->if_flags |= IFF_UP;
1517 			if ((error = (*ifp->if_init)(ifp)) != 0)
1518 				return error;
1519 		}
1520 #ifdef INET
1521 		if (ifa->ifa_addr->sa_family == AF_INET)
1522 			arp_ifinit(ifp, ifa);
1523 #endif
1524 		return 0;
1525 	    }
1526 
1527 	case SIOCSIFMTU:
1528 	    {
1529 		int maxmtu;
1530 
1531 		if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU)
1532 			maxmtu = ETHERMTU_JUMBO;
1533 		else
1534 			maxmtu = ETHERMTU;
1535 
1536 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu)
1537 			return EINVAL;
1538 		else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET)
1539 			return error;
1540 		else if (ifp->if_flags & IFF_UP) {
1541 			/* Make sure the device notices the MTU change. */
1542 			return (*ifp->if_init)(ifp);
1543 		} else
1544 			return 0;
1545 	    }
1546 
1547 	case SIOCSIFFLAGS:
1548 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
1549 			return error;
1550 		return ether_ioctl_reinit(ec);
1551 	case SIOCGIFFLAGS:
1552 		error = ifioctl_common(ifp, cmd, data);
1553 		if (error == 0) {
1554 			/* Set IFF_ALLMULTI for backcompat */
1555 			ifr->ifr_flags |= (ec->ec_flags & ETHER_F_ALLMULTI) ?
1556 			    IFF_ALLMULTI : 0;
1557 		}
1558 		return error;
1559 	case SIOCGETHERCAP:
1560 		eccr = (struct eccapreq *)data;
1561 		eccr->eccr_capabilities = ec->ec_capabilities;
1562 		eccr->eccr_capenable = ec->ec_capenable;
1563 		return 0;
1564 	case SIOCSETHERCAP:
1565 		eccr = (struct eccapreq *)data;
1566 		if ((eccr->eccr_capenable & ~ec->ec_capabilities) != 0)
1567 			return EINVAL;
1568 		if (eccr->eccr_capenable == ec->ec_capenable)
1569 			return 0;
1570 #if 0 /* notyet */
1571 		ec->ec_capenable = (ec->ec_capenable & ETHERCAP_CANTCHANGE)
1572 		    | (eccr->eccr_capenable & ~ETHERCAP_CANTCHANGE);
1573 #else
1574 		ec->ec_capenable = eccr->eccr_capenable;
1575 #endif
1576 		return ether_ioctl_reinit(ec);
1577 	case SIOCADDMULTI:
1578 		return ether_addmulti(ifreq_getaddr(cmd, ifr), ec);
1579 	case SIOCDELMULTI:
1580 		return ether_delmulti(ifreq_getaddr(cmd, ifr), ec);
1581 	case SIOCSIFMEDIA:
1582 	case SIOCGIFMEDIA:
1583 		if (ec->ec_mii != NULL)
1584 			return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media,
1585 			    cmd);
1586 		else if (ec->ec_ifmedia != NULL)
1587 			return ifmedia_ioctl(ifp, ifr, ec->ec_ifmedia, cmd);
1588 		else
1589 			return ENOTTY;
1590 		break;
1591 	case SIOCALIFADDR:
1592 		sdl = satocsdl(sstocsa(&iflr->addr));
1593 		if (sdl->sdl_family != AF_LINK)
1594 			;
1595 		else if (ETHER_IS_MULTICAST(CLLADDR(sdl)))
1596 			return EINVAL;
1597 		else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0)
1598 			return EINVAL;
1599 		/*FALLTHROUGH*/
1600 	default:
1601 		return ifioctl_common(ifp, cmd, data);
1602 	}
1603 	return 0;
1604 }
1605 
1606 /*
1607  * Enable/disable passing VLAN packets if the parent interface supports it.
1608  * Return:
1609  * 	 0: Ok
1610  *	-1: Parent interface does not support vlans
1611  *	>0: Error
1612  */
1613 int
1614 ether_enable_vlan_mtu(struct ifnet *ifp)
1615 {
1616 	int error;
1617 	struct ethercom *ec = (void *)ifp;
1618 
1619 	/* Parent does not support VLAN's */
1620 	if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0)
1621 		return -1;
1622 
1623 	/*
1624 	 * Parent supports the VLAN_MTU capability,
1625 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames;
1626 	 * enable it.
1627 	 */
1628 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1629 
1630 	/* Interface is down, defer for later */
1631 	if ((ifp->if_flags & IFF_UP) == 0)
1632 		return 0;
1633 
1634 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1635 		return 0;
1636 
1637 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1638 	return error;
1639 }
1640 
1641 int
1642 ether_disable_vlan_mtu(struct ifnet *ifp)
1643 {
1644 	int error;
1645 	struct ethercom *ec = (void *)ifp;
1646 
1647 	/* We still have VLAN's, defer for later */
1648 	if (ec->ec_nvlans != 0)
1649 		return 0;
1650 
1651 	/* Parent does not support VLAB's, nothing to do. */
1652 	if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0)
1653 		return -1;
1654 
1655 	/*
1656 	 * Disable Tx/Rx of VLAN-sized frames.
1657 	 */
1658 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1659 
1660 	/* Interface is down, defer for later */
1661 	if ((ifp->if_flags & IFF_UP) == 0)
1662 		return 0;
1663 
1664 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1665 		return 0;
1666 
1667 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1668 	return error;
1669 }
1670 
1671 /*
1672  * Add and delete VLAN TAG
1673  */
1674 int
1675 ether_add_vlantag(struct ifnet *ifp, uint16_t vtag, bool *vlanmtu_status)
1676 {
1677 	struct ethercom *ec = (void *)ifp;
1678 	struct vlanid_list *vidp;
1679 	bool vlanmtu_enabled;
1680 	uint16_t vid = EVL_VLANOFTAG(vtag);
1681 	int error;
1682 
1683 	vlanmtu_enabled = false;
1684 
1685 	/* Add a vid to the list */
1686 	vidp = kmem_alloc(sizeof(*vidp), KM_SLEEP);
1687 	vidp->vid = vid;
1688 
1689 	ETHER_LOCK(ec);
1690 	ec->ec_nvlans++;
1691 	SIMPLEQ_INSERT_TAIL(&ec->ec_vids, vidp, vid_list);
1692 	ETHER_UNLOCK(ec);
1693 
1694 	if (ec->ec_nvlans == 1) {
1695 		IFNET_LOCK(ifp);
1696 		error = ether_enable_vlan_mtu(ifp);
1697 		IFNET_UNLOCK(ifp);
1698 
1699 		if (error == 0) {
1700 			vlanmtu_enabled = true;
1701 		} else if (error != -1) {
1702 			goto fail;
1703 		}
1704 	}
1705 
1706 	if (ec->ec_vlan_cb != NULL) {
1707 		error = (*ec->ec_vlan_cb)(ec, vid, true);
1708 		if (error != 0)
1709 			goto fail;
1710 	}
1711 
1712 	if (vlanmtu_status != NULL)
1713 		*vlanmtu_status = vlanmtu_enabled;
1714 
1715 	return 0;
1716 fail:
1717 	ETHER_LOCK(ec);
1718 	ec->ec_nvlans--;
1719 	SIMPLEQ_REMOVE(&ec->ec_vids, vidp, vlanid_list, vid_list);
1720 	ETHER_UNLOCK(ec);
1721 
1722 	if (vlanmtu_enabled) {
1723 		IFNET_LOCK(ifp);
1724 		(void)ether_disable_vlan_mtu(ifp);
1725 		IFNET_UNLOCK(ifp);
1726 	}
1727 
1728 	kmem_free(vidp, sizeof(*vidp));
1729 
1730 	return error;
1731 }
1732 
1733 int
1734 ether_del_vlantag(struct ifnet *ifp, uint16_t vtag)
1735 {
1736 	struct ethercom *ec = (void *)ifp;
1737 	struct vlanid_list *vidp;
1738 	uint16_t vid = EVL_VLANOFTAG(vtag);
1739 
1740 	ETHER_LOCK(ec);
1741 	SIMPLEQ_FOREACH(vidp, &ec->ec_vids, vid_list) {
1742 		if (vidp->vid == vid) {
1743 			SIMPLEQ_REMOVE(&ec->ec_vids, vidp,
1744 			    vlanid_list, vid_list);
1745 			ec->ec_nvlans--;
1746 			break;
1747 		}
1748 	}
1749 	ETHER_UNLOCK(ec);
1750 
1751 	if (vidp == NULL)
1752 		return ENOENT;
1753 
1754 	if (ec->ec_vlan_cb != NULL) {
1755 		(void)(*ec->ec_vlan_cb)(ec, vidp->vid, false);
1756 	}
1757 
1758 	if (ec->ec_nvlans == 0) {
1759 		IFNET_LOCK(ifp);
1760 		(void)ether_disable_vlan_mtu(ifp);
1761 		IFNET_UNLOCK(ifp);
1762 	}
1763 
1764 	kmem_free(vidp, sizeof(*vidp));
1765 
1766 	return 0;
1767 }
1768 
1769 static int
1770 ether_multicast_sysctl(SYSCTLFN_ARGS)
1771 {
1772 	struct ether_multi *enm;
1773 	struct ifnet *ifp;
1774 	struct ethercom *ec;
1775 	int error = 0;
1776 	size_t written;
1777 	struct psref psref;
1778 	int bound;
1779 	unsigned int multicnt;
1780 	struct ether_multi_sysctl *addrs;
1781 	int i;
1782 
1783 	if (namelen != 1)
1784 		return EINVAL;
1785 
1786 	bound = curlwp_bind();
1787 	ifp = if_get_byindex(name[0], &psref);
1788 	if (ifp == NULL) {
1789 		error = ENODEV;
1790 		goto out;
1791 	}
1792 	if (ifp->if_type != IFT_ETHER) {
1793 		if_put(ifp, &psref);
1794 		*oldlenp = 0;
1795 		goto out;
1796 	}
1797 	ec = (struct ethercom *)ifp;
1798 
1799 	if (oldp == NULL) {
1800 		if_put(ifp, &psref);
1801 		*oldlenp = ec->ec_multicnt * sizeof(*addrs);
1802 		goto out;
1803 	}
1804 
1805 	/*
1806 	 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which
1807 	 * is sleepable, while holding it. Copy data to a local buffer first
1808 	 * with the lock taken and then call sysctl_copyout without holding it.
1809 	 */
1810 retry:
1811 	multicnt = ec->ec_multicnt;
1812 
1813 	if (multicnt == 0) {
1814 		if_put(ifp, &psref);
1815 		*oldlenp = 0;
1816 		goto out;
1817 	}
1818 
1819 	addrs = kmem_zalloc(sizeof(*addrs) * multicnt, KM_SLEEP);
1820 
1821 	ETHER_LOCK(ec);
1822 	if (multicnt != ec->ec_multicnt) {
1823 		/* The number of multicast addresses has changed */
1824 		ETHER_UNLOCK(ec);
1825 		kmem_free(addrs, sizeof(*addrs) * multicnt);
1826 		goto retry;
1827 	}
1828 
1829 	i = 0;
1830 	LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) {
1831 		struct ether_multi_sysctl *addr = &addrs[i];
1832 		addr->enm_refcount = enm->enm_refcount;
1833 		memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN);
1834 		memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN);
1835 		i++;
1836 	}
1837 	ETHER_UNLOCK(ec);
1838 
1839 	error = 0;
1840 	written = 0;
1841 	for (i = 0; i < multicnt; i++) {
1842 		struct ether_multi_sysctl *addr = &addrs[i];
1843 
1844 		if (written + sizeof(*addr) > *oldlenp)
1845 			break;
1846 		error = sysctl_copyout(l, addr, oldp, sizeof(*addr));
1847 		if (error)
1848 			break;
1849 		written += sizeof(*addr);
1850 		oldp = (char *)oldp + sizeof(*addr);
1851 	}
1852 	kmem_free(addrs, sizeof(*addrs) * multicnt);
1853 
1854 	if_put(ifp, &psref);
1855 
1856 	*oldlenp = written;
1857 out:
1858 	curlwp_bindx(bound);
1859 	return error;
1860 }
1861 
1862 static void
1863 ether_sysctl_setup(struct sysctllog **clog)
1864 {
1865 	const struct sysctlnode *rnode = NULL;
1866 
1867 	sysctl_createv(clog, 0, NULL, &rnode,
1868 		       CTLFLAG_PERMANENT,
1869 		       CTLTYPE_NODE, "ether",
1870 		       SYSCTL_DESCR("Ethernet-specific information"),
1871 		       NULL, 0, NULL, 0,
1872 		       CTL_NET, CTL_CREATE, CTL_EOL);
1873 
1874 	sysctl_createv(clog, 0, &rnode, NULL,
1875 		       CTLFLAG_PERMANENT,
1876 		       CTLTYPE_NODE, "multicast",
1877 		       SYSCTL_DESCR("multicast addresses"),
1878 		       ether_multicast_sysctl, 0, NULL, 0,
1879 		       CTL_CREATE, CTL_EOL);
1880 
1881 	sysctl_createv(clog, 0, &rnode, NULL,
1882 		       CTLFLAG_PERMANENT | CTLFLAG_READWRITE,
1883 		       CTLTYPE_STRING, "rps_hash",
1884 		       SYSCTL_DESCR("Interface rps hash function control"),
1885 		       sysctl_pktq_rps_hash_handler, 0, (void *)&ether_pktq_rps_hash_p,
1886 		       PKTQ_RPS_HASH_NAME_LEN,
1887 		       CTL_CREATE, CTL_EOL);
1888 }
1889 
1890 void
1891 etherinit(void)
1892 {
1893 
1894 #ifdef DIAGNOSTIC
1895 	mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET);
1896 #endif
1897 	ether_pktq_rps_hash_p = pktq_rps_hash_default;
1898 	ether_sysctl_setup(NULL);
1899 }
1900