xref: /netbsd-src/sys/net/if_ethersubr.c (revision dd3ee07da436799d8de85f3055253118b76bf345)
1 /*	$NetBSD: if_ethersubr.c,v 1.311 2022/04/04 06:10:00 yamaguchi 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 1.311 2022/04/04 06:10:00 yamaguchi 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 #if NBRIDGE > 0
129 #include <net/if_bridgevar.h>
130 #endif
131 
132 #include <netinet/in.h>
133 #ifdef INET
134 #include <netinet/in_var.h>
135 #endif
136 #include <netinet/if_inarp.h>
137 
138 #ifdef INET6
139 #ifndef INET
140 #include <netinet/in.h>
141 #endif
142 #include <netinet6/in6_var.h>
143 #include <netinet6/nd6.h>
144 #endif
145 
146 #include "carp.h"
147 #if NCARP > 0
148 #include <netinet/ip_carp.h>
149 #endif
150 
151 #ifdef NETATALK
152 #include <netatalk/at.h>
153 #include <netatalk/at_var.h>
154 #include <netatalk/at_extern.h>
155 
156 #define llc_snap_org_code llc_un.type_snap.org_code
157 #define llc_snap_ether_type llc_un.type_snap.ether_type
158 
159 extern u_char	at_org_code[3];
160 extern u_char	aarp_org_code[3];
161 #endif /* NETATALK */
162 
163 #ifdef MPLS
164 #include <netmpls/mpls.h>
165 #include <netmpls/mpls_var.h>
166 #endif
167 
168 CTASSERT(sizeof(struct ether_addr) == 6);
169 CTASSERT(sizeof(struct ether_header) == 14);
170 
171 #ifdef DIAGNOSTIC
172 static struct timeval bigpktppslim_last;
173 static int bigpktppslim = 2;	/* XXX */
174 static int bigpktpps_count;
175 static kmutex_t bigpktpps_lock __cacheline_aligned;
176 #endif
177 
178 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] =
179     { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
180 const uint8_t ethermulticastaddr_slowprotocols[ETHER_ADDR_LEN] =
181     { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x02 };
182 #define senderr(e) { error = (e); goto bad;}
183 
184 static pktq_rps_hash_func_t ether_pktq_rps_hash_p;
185 
186 static int ether_output(struct ifnet *, struct mbuf *,
187     const struct sockaddr *, const struct rtentry *);
188 
189 /*
190  * Ethernet output routine.
191  * Encapsulate a packet of type family for the local net.
192  * Assumes that ifp is actually pointer to ethercom structure.
193  */
194 static int
195 ether_output(struct ifnet * const ifp0, struct mbuf * const m0,
196     const struct sockaddr * const dst, const struct rtentry *rt)
197 {
198 	uint8_t esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
199 	uint16_t etype = 0;
200 	int error = 0, hdrcmplt = 0;
201 	struct mbuf *m = m0;
202 	struct mbuf *mcopy = NULL;
203 	struct ether_header *eh;
204 	struct ifnet *ifp = ifp0;
205 #ifdef INET
206 	struct arphdr *ah;
207 #endif
208 #ifdef NETATALK
209 	struct at_ifaddr *aa;
210 #endif
211 
212 #ifdef MBUFTRACE
213 	m_claimm(m, ifp->if_mowner);
214 #endif
215 
216 #if NCARP > 0
217 	if (ifp->if_type == IFT_CARP) {
218 		struct ifaddr *ifa;
219 		int s = pserialize_read_enter();
220 
221 		/* loop back if this is going to the carp interface */
222 		if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP &&
223 		    (ifa = ifa_ifwithaddr(dst)) != NULL) {
224 			if (ifa->ifa_ifp == ifp0) {
225 				pserialize_read_exit(s);
226 				return looutput(ifp0, m, dst, rt);
227 			}
228 		}
229 		pserialize_read_exit(s);
230 
231 		ifp = ifp->if_carpdev;
232 		/* ac = (struct arpcom *)ifp; */
233 
234 		if ((ifp0->if_flags & (IFF_UP | IFF_RUNNING)) !=
235 		    (IFF_UP | IFF_RUNNING))
236 			senderr(ENETDOWN);
237 	}
238 #endif
239 
240 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
241 		senderr(ENETDOWN);
242 
243 	switch (dst->sa_family) {
244 
245 #ifdef INET
246 	case AF_INET:
247 		if (m->m_flags & M_BCAST) {
248 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
249 		} else if (m->m_flags & M_MCAST) {
250 			ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst);
251 		} else {
252 			error = arpresolve(ifp0, rt, m, dst, edst, sizeof(edst));
253 			if (error)
254 				return (error == EWOULDBLOCK) ? 0 : error;
255 		}
256 		/* If broadcasting on a simplex interface, loopback a copy */
257 		if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
258 			mcopy = m_copypacket(m, M_DONTWAIT);
259 		etype = htons(ETHERTYPE_IP);
260 		break;
261 
262 	case AF_ARP:
263 		ah = mtod(m, struct arphdr *);
264 		if (m->m_flags & M_BCAST) {
265 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
266 		} else {
267 			void *tha = ar_tha(ah);
268 
269 			if (tha == NULL) {
270 				/* fake with ARPHRD_IEEE1394 */
271 				m_freem(m);
272 				return 0;
273 			}
274 			memcpy(edst, tha, sizeof(edst));
275 		}
276 
277 		ah->ar_hrd = htons(ARPHRD_ETHER);
278 
279 		switch (ntohs(ah->ar_op)) {
280 		case ARPOP_REVREQUEST:
281 		case ARPOP_REVREPLY:
282 			etype = htons(ETHERTYPE_REVARP);
283 			break;
284 
285 		case ARPOP_REQUEST:
286 		case ARPOP_REPLY:
287 		default:
288 			etype = htons(ETHERTYPE_ARP);
289 		}
290 		break;
291 #endif
292 
293 #ifdef INET6
294 	case AF_INET6:
295 		if (m->m_flags & M_BCAST) {
296 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
297 		} else if (m->m_flags & M_MCAST) {
298 			ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr,
299 			    edst);
300 		} else {
301 			error = nd6_resolve(ifp0, rt, m, dst, edst,
302 			    sizeof(edst));
303 			if (error)
304 				return (error == EWOULDBLOCK) ? 0 : error;
305 		}
306 		etype = htons(ETHERTYPE_IPV6);
307 		break;
308 #endif
309 
310 #ifdef NETATALK
311 	case AF_APPLETALK: {
312 		struct ifaddr *ifa;
313 		int s;
314 
315 		KERNEL_LOCK(1, NULL);
316 
317 		if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) {
318 			KERNEL_UNLOCK_ONE(NULL);
319 			return 0;
320 		}
321 
322 		/*
323 		 * ifaddr is the first thing in at_ifaddr
324 		 */
325 		s = pserialize_read_enter();
326 		ifa = at_ifawithnet((const struct sockaddr_at *)dst, ifp);
327 		if (ifa == NULL) {
328 			pserialize_read_exit(s);
329 			KERNEL_UNLOCK_ONE(NULL);
330 			senderr(EADDRNOTAVAIL);
331 		}
332 		aa = (struct at_ifaddr *)ifa;
333 
334 		/*
335 		 * In the phase 2 case, we need to prepend an mbuf for the
336 		 * llc header.
337 		 */
338 		if (aa->aa_flags & AFA_PHASE2) {
339 			struct llc llc;
340 
341 			M_PREPEND(m, sizeof(struct llc), M_DONTWAIT);
342 			if (m == NULL) {
343 				pserialize_read_exit(s);
344 				KERNEL_UNLOCK_ONE(NULL);
345 				senderr(ENOBUFS);
346 			}
347 
348 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
349 			llc.llc_control = LLC_UI;
350 			memcpy(llc.llc_snap_org_code, at_org_code,
351 			    sizeof(llc.llc_snap_org_code));
352 			llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK);
353 			memcpy(mtod(m, void *), &llc, sizeof(struct llc));
354 		} else {
355 			etype = htons(ETHERTYPE_ATALK);
356 		}
357 		pserialize_read_exit(s);
358 		KERNEL_UNLOCK_ONE(NULL);
359 		break;
360 	}
361 #endif /* NETATALK */
362 
363 	case pseudo_AF_HDRCMPLT:
364 		hdrcmplt = 1;
365 		memcpy(esrc,
366 		    ((const struct ether_header *)dst->sa_data)->ether_shost,
367 		    sizeof(esrc));
368 		/* FALLTHROUGH */
369 
370 	case AF_UNSPEC:
371 		memcpy(edst,
372 		    ((const struct ether_header *)dst->sa_data)->ether_dhost,
373 		    sizeof(edst));
374 		/* AF_UNSPEC doesn't swap the byte order of the ether_type. */
375 		etype = ((const struct ether_header *)dst->sa_data)->ether_type;
376 		break;
377 
378 	default:
379 		printf("%s: can't handle af%d\n", ifp->if_xname,
380 		    dst->sa_family);
381 		senderr(EAFNOSUPPORT);
382 	}
383 
384 #ifdef MPLS
385 	{
386 		struct m_tag *mtag;
387 		mtag = m_tag_find(m, PACKET_TAG_MPLS);
388 		if (mtag != NULL) {
389 			/* Having the tag itself indicates it's MPLS */
390 			etype = htons(ETHERTYPE_MPLS);
391 			m_tag_delete(m, mtag);
392 		}
393 	}
394 #endif
395 
396 	if (mcopy)
397 		(void)looutput(ifp, mcopy, dst, rt);
398 
399 	KASSERT((m->m_flags & M_PKTHDR) != 0);
400 
401 	/*
402 	 * If no ether type is set, this must be a 802.2 formatted packet.
403 	 */
404 	if (etype == 0)
405 		etype = htons(m->m_pkthdr.len);
406 
407 	/*
408 	 * Add local net header. If no space in first mbuf, allocate another.
409 	 */
410 	M_PREPEND(m, sizeof(struct ether_header), M_DONTWAIT);
411 	if (m == NULL)
412 		senderr(ENOBUFS);
413 
414 	eh = mtod(m, struct ether_header *);
415 	/* Note: etype is already in network byte order. */
416 	memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
417 	memcpy(eh->ether_dhost, edst, sizeof(edst));
418 	if (hdrcmplt) {
419 		memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost));
420 	} else {
421 	 	memcpy(eh->ether_shost, CLLADDR(ifp->if_sadl),
422 		    sizeof(eh->ether_shost));
423 	}
424 
425 #if NCARP > 0
426 	if (ifp0 != ifp && ifp0->if_type == IFT_CARP) {
427 	 	memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl),
428 		    sizeof(eh->ether_shost));
429 	}
430 #endif
431 
432 	if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0)
433 		return error;
434 	if (m == NULL)
435 		return 0;
436 
437 #if NBRIDGE > 0
438 	/*
439 	 * Bridges require special output handling.
440 	 */
441 	if (ifp->if_bridge)
442 		return bridge_output(ifp, m, NULL, NULL);
443 #endif
444 
445 #if NCARP > 0
446 	if (ifp != ifp0)
447 		if_statadd(ifp0, if_obytes, m->m_pkthdr.len + ETHER_HDR_LEN);
448 #endif
449 
450 #ifdef ALTQ
451 	KERNEL_LOCK(1, NULL);
452 	/*
453 	 * If ALTQ is enabled on the parent interface, do
454 	 * classification; the queueing discipline might not
455 	 * require classification, but might require the
456 	 * address family/header pointer in the pktattr.
457 	 */
458 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
459 		altq_etherclassify(&ifp->if_snd, m);
460 	KERNEL_UNLOCK_ONE(NULL);
461 #endif
462 	return ifq_enqueue(ifp, m);
463 
464 bad:
465 	if_statinc(ifp, if_oerrors);
466 	if (m)
467 		m_freem(m);
468 	return error;
469 }
470 
471 #ifdef ALTQ
472 /*
473  * This routine is a slight hack to allow a packet to be classified
474  * if the Ethernet headers are present.  It will go away when ALTQ's
475  * classification engine understands link headers.
476  *
477  * XXX: We may need to do m_pullups here. First to ensure struct ether_header
478  * is indeed contiguous, then to read the LLC and so on.
479  */
480 void
481 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m)
482 {
483 	struct ether_header *eh;
484 	struct mbuf *mtop = m;
485 	uint16_t ether_type;
486 	int hlen, af, hdrsize;
487 	void *hdr;
488 
489 	KASSERT((mtop->m_flags & M_PKTHDR) != 0);
490 
491 	hlen = ETHER_HDR_LEN;
492 	eh = mtod(m, struct ether_header *);
493 
494 	ether_type = htons(eh->ether_type);
495 
496 	if (ether_type < ETHERMTU) {
497 		/* LLC/SNAP */
498 		struct llc *llc = (struct llc *)(eh + 1);
499 		hlen += 8;
500 
501 		if (m->m_len < hlen ||
502 		    llc->llc_dsap != LLC_SNAP_LSAP ||
503 		    llc->llc_ssap != LLC_SNAP_LSAP ||
504 		    llc->llc_control != LLC_UI) {
505 			/* Not SNAP. */
506 			goto bad;
507 		}
508 
509 		ether_type = htons(llc->llc_un.type_snap.ether_type);
510 	}
511 
512 	switch (ether_type) {
513 	case ETHERTYPE_IP:
514 		af = AF_INET;
515 		hdrsize = 20;		/* sizeof(struct ip) */
516 		break;
517 
518 	case ETHERTYPE_IPV6:
519 		af = AF_INET6;
520 		hdrsize = 40;		/* sizeof(struct ip6_hdr) */
521 		break;
522 
523 	default:
524 		af = AF_UNSPEC;
525 		hdrsize = 0;
526 		break;
527 	}
528 
529 	while (m->m_len <= hlen) {
530 		hlen -= m->m_len;
531 		m = m->m_next;
532 		if (m == NULL)
533 			goto bad;
534 	}
535 
536 	if (m->m_len < (hlen + hdrsize)) {
537 		/*
538 		 * protocol header not in a single mbuf.
539 		 * We can't cope with this situation right
540 		 * now (but it shouldn't ever happen, really, anyhow).
541 		 */
542 #ifdef DEBUG
543 		printf("altq_etherclassify: headers span multiple mbufs: "
544 		    "%d < %d\n", m->m_len, (hlen + hdrsize));
545 #endif
546 		goto bad;
547 	}
548 
549 	m->m_data += hlen;
550 	m->m_len -= hlen;
551 
552 	hdr = mtod(m, void *);
553 
554 	if (ALTQ_NEEDS_CLASSIFY(ifq)) {
555 		mtop->m_pkthdr.pattr_class =
556 		    (*ifq->altq_classify)(ifq->altq_clfier, m, af);
557 	}
558 	mtop->m_pkthdr.pattr_af = af;
559 	mtop->m_pkthdr.pattr_hdr = hdr;
560 
561 	m->m_data -= hlen;
562 	m->m_len += hlen;
563 
564 	return;
565 
566 bad:
567 	mtop->m_pkthdr.pattr_class = NULL;
568 	mtop->m_pkthdr.pattr_hdr = NULL;
569 	mtop->m_pkthdr.pattr_af = AF_UNSPEC;
570 }
571 #endif /* ALTQ */
572 
573 #if defined (LLC) || defined (NETATALK)
574 static void
575 ether_input_llc(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh)
576 {
577 	struct ifqueue *inq = NULL;
578 	int isr = 0;
579 	struct llc *l;
580 
581 	if (m->m_len < sizeof(*eh) + sizeof(struct llc))
582 		goto error;
583 
584 	l = (struct llc *)(eh+1);
585 	switch (l->llc_dsap) {
586 #ifdef NETATALK
587 	case LLC_SNAP_LSAP:
588 		switch (l->llc_control) {
589 		case LLC_UI:
590 			if (l->llc_ssap != LLC_SNAP_LSAP)
591 				goto error;
592 
593 			if (memcmp(&(l->llc_snap_org_code)[0],
594 			    at_org_code, sizeof(at_org_code)) == 0 &&
595 			    ntohs(l->llc_snap_ether_type) ==
596 			    ETHERTYPE_ATALK) {
597 				inq = &atintrq2;
598 				m_adj(m, sizeof(struct ether_header)
599 				    + sizeof(struct llc));
600 				isr = NETISR_ATALK;
601 				break;
602 			}
603 
604 			if (memcmp(&(l->llc_snap_org_code)[0],
605 			    aarp_org_code,
606 			    sizeof(aarp_org_code)) == 0 &&
607 			    ntohs(l->llc_snap_ether_type) ==
608 			    ETHERTYPE_AARP) {
609 				m_adj(m, sizeof(struct ether_header)
610 				    + sizeof(struct llc));
611 				aarpinput(ifp, m); /* XXX queue? */
612 				return;
613 			}
614 
615 		default:
616 			goto error;
617 		}
618 		break;
619 #endif
620 	default:
621 		goto noproto;
622 	}
623 
624 	KASSERT(inq != NULL);
625 	IFQ_ENQUEUE_ISR(inq, m, isr);
626 	return;
627 
628 noproto:
629 	m_freem(m);
630 	if_statinc(ifp, if_noproto);
631 	return;
632 error:
633 	m_freem(m);
634 	if_statinc(ifp, if_ierrors);
635 	return;
636 }
637 #endif /* defined (LLC) || defined (NETATALK) */
638 
639 /*
640  * Process a received Ethernet packet;
641  * the packet is in the mbuf chain m with
642  * the ether header.
643  */
644 void
645 ether_input(struct ifnet *ifp, struct mbuf *m)
646 {
647 	struct ethercom *ec = (struct ethercom *) ifp;
648 	pktqueue_t *pktq = NULL;
649 	struct ifqueue *inq = NULL;
650 	uint16_t etype;
651 	struct ether_header *eh;
652 	size_t ehlen;
653 	static int earlypkts;
654 	int isr = 0;
655 
656 	KASSERT(!cpu_intr_p());
657 	KASSERT((m->m_flags & M_PKTHDR) != 0);
658 
659 	if ((ifp->if_flags & IFF_UP) == 0)
660 		goto drop;
661 
662 #ifdef MBUFTRACE
663 	m_claimm(m, &ec->ec_rx_mowner);
664 #endif
665 
666 	if (__predict_false(m->m_len < sizeof(*eh))) {
667 		if ((m = m_pullup(m, sizeof(*eh))) == NULL) {
668 			if_statinc(ifp, if_ierrors);
669 			return;
670 		}
671 	}
672 
673 	eh = mtod(m, struct ether_header *);
674 	etype = ntohs(eh->ether_type);
675 	ehlen = sizeof(*eh);
676 
677 	if (__predict_false(earlypkts < 100 ||
678 		entropy_epoch() == (unsigned)-1)) {
679 		rnd_add_data(NULL, eh, ehlen, 0);
680 		earlypkts++;
681 	}
682 
683 	/*
684 	 * Determine if the packet is within its size limits. For MPLS the
685 	 * header length is variable, so we skip the check.
686 	 */
687 	if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len >
688 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
689 #ifdef DIAGNOSTIC
690 		mutex_enter(&bigpktpps_lock);
691 		if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count,
692 		    bigpktppslim)) {
693 			printf("%s: discarding oversize frame (len=%d)\n",
694 			    ifp->if_xname, m->m_pkthdr.len);
695 		}
696 		mutex_exit(&bigpktpps_lock);
697 #endif
698 		goto error;
699 	}
700 
701 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
702 		/*
703 		 * If this is not a simplex interface, drop the packet
704 		 * if it came from us.
705 		 */
706 		if ((ifp->if_flags & IFF_SIMPLEX) == 0 &&
707 		    memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost,
708 		    ETHER_ADDR_LEN) == 0) {
709 			goto drop;
710 		}
711 
712 		if (memcmp(etherbroadcastaddr,
713 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
714 			m->m_flags |= M_BCAST;
715 		else
716 			m->m_flags |= M_MCAST;
717 		if_statinc(ifp, if_imcasts);
718 	}
719 
720 	/* If the CRC is still on the packet, trim it off. */
721 	if (m->m_flags & M_HASFCS) {
722 		m_adj(m, -ETHER_CRC_LEN);
723 		m->m_flags &= ~M_HASFCS;
724 	}
725 
726 	if_statadd(ifp, if_ibytes, m->m_pkthdr.len);
727 
728 #if NCARP > 0
729 	if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) {
730 		/*
731 		 * Clear M_PROMISC, in case the packet comes from a
732 		 * vlan.
733 		 */
734 		m->m_flags &= ~M_PROMISC;
735 		if (carp_input(m, (uint8_t *)&eh->ether_shost,
736 		    (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0)
737 			return;
738 	}
739 #endif
740 
741 	if ((m->m_flags & (M_BCAST | M_MCAST | M_PROMISC)) == 0 &&
742 	    (ifp->if_flags & IFF_PROMISC) != 0 &&
743 	    memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost,
744 	     ETHER_ADDR_LEN) != 0) {
745 		m->m_flags |= M_PROMISC;
746 	}
747 
748 	if ((m->m_flags & M_PROMISC) == 0) {
749 		if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
750 			return;
751 		if (m == NULL)
752 			return;
753 
754 		eh = mtod(m, struct ether_header *);
755 		etype = ntohs(eh->ether_type);
756 	}
757 
758 #if NAGR > 0
759 	if (ifp->if_lagg != NULL &&
760 	    __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) {
761 		m->m_flags &= ~M_PROMISC;
762 		agr_input(ifp, m);
763 		return;
764 	}
765 #endif
766 
767 	/*
768 	 * If VLANs are configured on the interface, check to
769 	 * see if the device performed the decapsulation and
770 	 * provided us with the tag.
771 	 */
772 	if (ec->ec_nvlans && vlan_has_tag(m)) {
773 #if NVLAN > 0
774 		/*
775 		 * vlan_input() will either recursively call ether_input()
776 		 * or drop the packet.
777 		 */
778 		vlan_input(ifp, m);
779 		return;
780 #else
781 		goto noproto;
782 #endif
783 	}
784 
785 	/*
786 	 * Handle protocols that expect to have the Ethernet header
787 	 * (and possibly FCS) intact.
788 	 */
789 	switch (etype) {
790 	case ETHERTYPE_VLAN: {
791 		struct ether_vlan_header *evl = (void *)eh;
792 
793 		/*
794 		 * If there is a tag of 0, then the VLAN header was probably
795 		 * just being used to store the priority.  Extract the ether
796 		 * type, and if IP or IPV6, let them deal with it.
797 		 */
798 		if (m->m_len >= sizeof(*evl) &&
799 		    EVL_VLANOFTAG(ntohs(evl->evl_tag)) == 0) {
800 			etype = ntohs(evl->evl_proto);
801 			ehlen = sizeof(*evl);
802 			if ((m->m_flags & M_PROMISC) == 0 &&
803 			    (etype == ETHERTYPE_IP ||
804 			     etype == ETHERTYPE_IPV6))
805 				break;
806 		}
807 
808 #if NVLAN > 0
809 		/*
810 		 * vlan_input() will either recursively call ether_input()
811 		 * or drop the packet.
812 		 */
813 		if (ec->ec_nvlans != 0) {
814 			vlan_input(ifp, m);
815 			return;
816 		} else
817 #endif
818 			goto noproto;
819 	}
820 
821 #if NPPPOE > 0
822 	case ETHERTYPE_PPPOEDISC:
823 		pppoedisc_input(ifp, m);
824 		return;
825 
826 	case ETHERTYPE_PPPOE:
827 		pppoe_input(ifp, m);
828 		return;
829 #endif
830 
831 	case ETHERTYPE_SLOWPROTOCOLS: {
832 		uint8_t subtype;
833 
834 		if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype))
835 			goto error;
836 
837 		m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype);
838 		switch (subtype) {
839 #if NAGR > 0
840 		case SLOWPROTOCOLS_SUBTYPE_LACP:
841 			if (ifp->if_lagg != NULL) {
842 				ieee8023ad_lacp_input(ifp, m);
843 				return;
844 			}
845 			break;
846 
847 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
848 			if (ifp->if_lagg != NULL) {
849 				ieee8023ad_marker_input(ifp, m);
850 				return;
851 			}
852 			break;
853 #endif
854 
855 		default:
856 			if (subtype == 0 || subtype > 10) {
857 				/* illegal value */
858 				goto error;
859 			}
860 			/* unknown subtype */
861 			break;
862 		}
863 	}
864 	/* FALLTHROUGH */
865 	default:
866 		if (m->m_flags & M_PROMISC)
867 			goto drop;
868 	}
869 
870 	/* If the CRC is still on the packet, trim it off. */
871 	if (m->m_flags & M_HASFCS) {
872 		m_adj(m, -ETHER_CRC_LEN);
873 		m->m_flags &= ~M_HASFCS;
874 	}
875 
876 	/* etype represents the size of the payload in this case */
877 	if (etype <= ETHERMTU + sizeof(struct ether_header)) {
878 		KASSERT(ehlen == sizeof(*eh));
879 #if defined (LLC) || defined (NETATALK)
880 		ether_input_llc(ifp, m, eh);
881 		return;
882 #else
883 		/* ethertype of 0-1500 is regarded as noproto */
884 		goto noproto;
885 #endif
886 	}
887 
888 	/* Strip off the Ethernet header. */
889 	m_adj(m, ehlen);
890 
891 	switch (etype) {
892 #ifdef INET
893 	case ETHERTYPE_IP:
894 #ifdef GATEWAY
895 		if (ipflow_fastforward(m))
896 			return;
897 #endif
898 		pktq = ip_pktq;
899 		break;
900 
901 	case ETHERTYPE_ARP:
902 		isr = NETISR_ARP;
903 		inq = &arpintrq;
904 		break;
905 
906 	case ETHERTYPE_REVARP:
907 		revarpinput(m);	/* XXX queue? */
908 		return;
909 #endif
910 
911 #ifdef INET6
912 	case ETHERTYPE_IPV6:
913 		if (__predict_false(!in6_present))
914 			goto noproto;
915 #ifdef GATEWAY
916 		if (ip6flow_fastforward(&m))
917 			return;
918 #endif
919 		pktq = ip6_pktq;
920 		break;
921 #endif
922 
923 #ifdef NETATALK
924 	case ETHERTYPE_ATALK:
925 		isr = NETISR_ATALK;
926 		inq = &atintrq1;
927 		break;
928 
929 	case ETHERTYPE_AARP:
930 		aarpinput(ifp, m); /* XXX queue? */
931 		return;
932 #endif
933 
934 #ifdef MPLS
935 	case ETHERTYPE_MPLS:
936 		isr = NETISR_MPLS;
937 		inq = &mplsintrq;
938 		break;
939 #endif
940 
941 	default:
942 		goto noproto;
943 	}
944 
945 	if (__predict_true(pktq)) {
946 		const uint32_t h = pktq_rps_hash(&ether_pktq_rps_hash_p, m);
947 		if (__predict_false(!pktq_enqueue(pktq, m, h))) {
948 			m_freem(m);
949 		}
950 		return;
951 	}
952 
953 	if (__predict_false(!inq)) {
954 		/* Should not happen. */
955 		goto error;
956 	}
957 
958 	IFQ_ENQUEUE_ISR(inq, m, isr);
959 	return;
960 
961 drop:
962 	m_freem(m);
963 	if_statinc(ifp, if_iqdrops);
964 	return;
965 noproto:
966 	m_freem(m);
967 	if_statinc(ifp, if_noproto);
968 	return;
969 error:
970 	m_freem(m);
971 	if_statinc(ifp, if_ierrors);
972 	return;
973 }
974 
975 /*
976  * Convert Ethernet address to printable (loggable) representation.
977  */
978 char *
979 ether_sprintf(const u_char *ap)
980 {
981 	static char etherbuf[3 * ETHER_ADDR_LEN];
982 	return ether_snprintf(etherbuf, sizeof(etherbuf), ap);
983 }
984 
985 char *
986 ether_snprintf(char *buf, size_t len, const u_char *ap)
987 {
988 	char *cp = buf;
989 	size_t i;
990 
991 	for (i = 0; i < len / 3; i++) {
992 		*cp++ = hexdigits[*ap >> 4];
993 		*cp++ = hexdigits[*ap++ & 0xf];
994 		*cp++ = ':';
995 	}
996 	*--cp = '\0';
997 	return buf;
998 }
999 
1000 /*
1001  * Perform common duties while attaching to interface list
1002  */
1003 void
1004 ether_ifattach(struct ifnet *ifp, const uint8_t *lla)
1005 {
1006 	struct ethercom *ec = (struct ethercom *)ifp;
1007 	char xnamebuf[HOOKNAMSIZ];
1008 
1009 	ifp->if_type = IFT_ETHER;
1010 	ifp->if_hdrlen = ETHER_HDR_LEN;
1011 	ifp->if_dlt = DLT_EN10MB;
1012 	ifp->if_mtu = ETHERMTU;
1013 	ifp->if_output = ether_output;
1014 	ifp->_if_input = ether_input;
1015 	if (ifp->if_baudrate == 0)
1016 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
1017 
1018 	if (lla != NULL)
1019 		if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla));
1020 
1021 	LIST_INIT(&ec->ec_multiaddrs);
1022 	SIMPLEQ_INIT(&ec->ec_vids);
1023 	ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
1024 	ec->ec_flags = 0;
1025 	ifp->if_broadcastaddr = etherbroadcastaddr;
1026 	bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header));
1027 	snprintf(xnamebuf, sizeof(xnamebuf),
1028 	    "%s-ether_ifdetachhooks", ifp->if_xname);
1029 	ec->ec_ifdetach_hooks = simplehook_create(IPL_NET, xnamebuf);
1030 #ifdef MBUFTRACE
1031 	mowner_init_owner(&ec->ec_tx_mowner, ifp->if_xname, "tx");
1032 	mowner_init_owner(&ec->ec_rx_mowner, ifp->if_xname, "rx");
1033 	MOWNER_ATTACH(&ec->ec_tx_mowner);
1034 	MOWNER_ATTACH(&ec->ec_rx_mowner);
1035 	ifp->if_mowner = &ec->ec_tx_mowner;
1036 #endif
1037 }
1038 
1039 void
1040 ether_ifdetach(struct ifnet *ifp)
1041 {
1042 	struct ethercom *ec = (void *) ifp;
1043 	struct ether_multi *enm;
1044 
1045 	IFNET_ASSERT_UNLOCKED(ifp);
1046 	/*
1047 	 * Prevent further calls to ioctl (for example turning off
1048 	 * promiscuous mode from the bridge code), which eventually can
1049 	 * call if_init() which can cause panics because the interface
1050 	 * is in the process of being detached. Return device not configured
1051 	 * instead.
1052 	 */
1053 	ifp->if_ioctl = __FPTRCAST(int (*)(struct ifnet *, u_long, void *),
1054 	    enxio);
1055 
1056 	simplehook_dohooks(ec->ec_ifdetach_hooks);
1057 	KASSERT(!simplehook_has_hooks(ec->ec_ifdetach_hooks));
1058 	simplehook_destroy(ec->ec_ifdetach_hooks);
1059 
1060 	bpf_detach(ifp);
1061 
1062 	ETHER_LOCK(ec);
1063 	KASSERT(ec->ec_nvlans == 0);
1064 	while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) {
1065 		LIST_REMOVE(enm, enm_list);
1066 		kmem_free(enm, sizeof(*enm));
1067 		ec->ec_multicnt--;
1068 	}
1069 	ETHER_UNLOCK(ec);
1070 
1071 	mutex_obj_free(ec->ec_lock);
1072 	ec->ec_lock = NULL;
1073 
1074 	ifp->if_mowner = NULL;
1075 	MOWNER_DETACH(&ec->ec_rx_mowner);
1076 	MOWNER_DETACH(&ec->ec_tx_mowner);
1077 }
1078 
1079 void *
1080 ether_ifdetachhook_establish(struct ifnet *ifp,
1081     void (*fn)(void *), void *arg)
1082 {
1083 	struct ethercom *ec;
1084 	khook_t *hk;
1085 
1086 	if (ifp->if_type != IFT_ETHER)
1087 		return NULL;
1088 
1089 	ec = (struct ethercom *)ifp;
1090 	hk = simplehook_establish(ec->ec_ifdetach_hooks,
1091 	    fn, arg);
1092 
1093 	return (void *)hk;
1094 }
1095 
1096 void
1097 ether_ifdetachhook_disestablish(struct ifnet *ifp,
1098     void *vhook, kmutex_t *lock)
1099 {
1100 	struct ethercom *ec;
1101 
1102 	if (vhook == NULL)
1103 		return;
1104 
1105 	ec = (struct ethercom *)ifp;
1106 	simplehook_disestablish(ec->ec_ifdetach_hooks, vhook, lock);
1107 }
1108 
1109 #if 0
1110 /*
1111  * This is for reference.  We have a table-driven version
1112  * of the little-endian crc32 generator, which is faster
1113  * than the double-loop.
1114  */
1115 uint32_t
1116 ether_crc32_le(const uint8_t *buf, size_t len)
1117 {
1118 	uint32_t c, crc, carry;
1119 	size_t i, j;
1120 
1121 	crc = 0xffffffffU;	/* initial value */
1122 
1123 	for (i = 0; i < len; i++) {
1124 		c = buf[i];
1125 		for (j = 0; j < 8; j++) {
1126 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1127 			crc >>= 1;
1128 			c >>= 1;
1129 			if (carry)
1130 				crc = (crc ^ ETHER_CRC_POLY_LE);
1131 		}
1132 	}
1133 
1134 	return (crc);
1135 }
1136 #else
1137 uint32_t
1138 ether_crc32_le(const uint8_t *buf, size_t len)
1139 {
1140 	static const uint32_t crctab[] = {
1141 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1142 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1143 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1144 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1145 	};
1146 	uint32_t crc;
1147 	size_t i;
1148 
1149 	crc = 0xffffffffU;	/* initial value */
1150 
1151 	for (i = 0; i < len; i++) {
1152 		crc ^= buf[i];
1153 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1154 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1155 	}
1156 
1157 	return (crc);
1158 }
1159 #endif
1160 
1161 uint32_t
1162 ether_crc32_be(const uint8_t *buf, size_t len)
1163 {
1164 	uint32_t c, crc, carry;
1165 	size_t i, j;
1166 
1167 	crc = 0xffffffffU;	/* initial value */
1168 
1169 	for (i = 0; i < len; i++) {
1170 		c = buf[i];
1171 		for (j = 0; j < 8; j++) {
1172 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1173 			crc <<= 1;
1174 			c >>= 1;
1175 			if (carry)
1176 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1177 		}
1178 	}
1179 
1180 	return (crc);
1181 }
1182 
1183 #ifdef INET
1184 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] =
1185     { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 };
1186 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] =
1187     { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff };
1188 #endif
1189 #ifdef INET6
1190 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] =
1191     { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 };
1192 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] =
1193     { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff };
1194 #endif
1195 
1196 /*
1197  * ether_aton implementation, not using a static buffer.
1198  */
1199 int
1200 ether_aton_r(u_char *dest, size_t len, const char *str)
1201 {
1202 	const u_char *cp = (const void *)str;
1203 	u_char *ep;
1204 
1205 #define atox(c)	(((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10))
1206 
1207 	if (len < ETHER_ADDR_LEN)
1208 		return ENOSPC;
1209 
1210 	ep = dest + ETHER_ADDR_LEN;
1211 
1212 	while (*cp) {
1213 		if (!isxdigit(*cp))
1214 			return EINVAL;
1215 
1216 		*dest = atox(*cp);
1217 		cp++;
1218 		if (isxdigit(*cp)) {
1219 			*dest = (*dest << 4) | atox(*cp);
1220 			cp++;
1221 		}
1222 		dest++;
1223 
1224 		if (dest == ep)
1225 			return (*cp == '\0') ? 0 : ENAMETOOLONG;
1226 
1227 		switch (*cp) {
1228 		case ':':
1229 		case '-':
1230 		case '.':
1231 			cp++;
1232 			break;
1233 		}
1234 	}
1235 	return ENOBUFS;
1236 }
1237 
1238 /*
1239  * Convert a sockaddr into an Ethernet address or range of Ethernet
1240  * addresses.
1241  */
1242 int
1243 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN],
1244     uint8_t addrhi[ETHER_ADDR_LEN])
1245 {
1246 #ifdef INET
1247 	const struct sockaddr_in *sin;
1248 #endif
1249 #ifdef INET6
1250 	const struct sockaddr_in6 *sin6;
1251 #endif
1252 
1253 	switch (sa->sa_family) {
1254 
1255 	case AF_UNSPEC:
1256 		memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN);
1257 		memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1258 		break;
1259 
1260 #ifdef INET
1261 	case AF_INET:
1262 		sin = satocsin(sa);
1263 		if (sin->sin_addr.s_addr == INADDR_ANY) {
1264 			/*
1265 			 * An IP address of INADDR_ANY means listen to
1266 			 * or stop listening to all of the Ethernet
1267 			 * multicast addresses used for IP.
1268 			 * (This is for the sake of IP multicast routers.)
1269 			 */
1270 			memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN);
1271 			memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN);
1272 		} else {
1273 			ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo);
1274 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1275 		}
1276 		break;
1277 #endif
1278 #ifdef INET6
1279 	case AF_INET6:
1280 		sin6 = satocsin6(sa);
1281 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1282 			/*
1283 			 * An IP6 address of 0 means listen to or stop
1284 			 * listening to all of the Ethernet multicast
1285 			 * address used for IP6.
1286 			 * (This is used for multicast routers.)
1287 			 */
1288 			memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN);
1289 			memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN);
1290 		} else {
1291 			ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo);
1292 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1293 		}
1294 		break;
1295 #endif
1296 
1297 	default:
1298 		return EAFNOSUPPORT;
1299 	}
1300 	return 0;
1301 }
1302 
1303 /*
1304  * Add an Ethernet multicast address or range of addresses to the list for a
1305  * given interface.
1306  */
1307 int
1308 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec)
1309 {
1310 	struct ether_multi *enm, *_enm;
1311 	u_char addrlo[ETHER_ADDR_LEN];
1312 	u_char addrhi[ETHER_ADDR_LEN];
1313 	int error = 0;
1314 
1315 	/* Allocate out of lock */
1316 	enm = kmem_alloc(sizeof(*enm), KM_SLEEP);
1317 
1318 	ETHER_LOCK(ec);
1319 	error = ether_multiaddr(sa, addrlo, addrhi);
1320 	if (error != 0)
1321 		goto out;
1322 
1323 	/*
1324 	 * Verify that we have valid Ethernet multicast addresses.
1325 	 */
1326 	if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) {
1327 		error = EINVAL;
1328 		goto out;
1329 	}
1330 
1331 	/*
1332 	 * See if the address range is already in the list.
1333 	 */
1334 	_enm = ether_lookup_multi(addrlo, addrhi, ec);
1335 	if (_enm != NULL) {
1336 		/*
1337 		 * Found it; just increment the reference count.
1338 		 */
1339 		++_enm->enm_refcount;
1340 		error = 0;
1341 		goto out;
1342 	}
1343 
1344 	/*
1345 	 * Link a new multicast record into the interface's multicast list.
1346 	 */
1347 	memcpy(enm->enm_addrlo, addrlo, ETHER_ADDR_LEN);
1348 	memcpy(enm->enm_addrhi, addrhi, ETHER_ADDR_LEN);
1349 	enm->enm_refcount = 1;
1350 	LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list);
1351 	ec->ec_multicnt++;
1352 
1353 	/*
1354 	 * Return ENETRESET to inform the driver that the list has changed
1355 	 * and its reception filter should be adjusted accordingly.
1356 	 */
1357 	error = ENETRESET;
1358 	enm = NULL;
1359 
1360 out:
1361 	ETHER_UNLOCK(ec);
1362 	if (enm != NULL)
1363 		kmem_free(enm, sizeof(*enm));
1364 	return error;
1365 }
1366 
1367 /*
1368  * Delete a multicast address record.
1369  */
1370 int
1371 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec)
1372 {
1373 	struct ether_multi *enm;
1374 	u_char addrlo[ETHER_ADDR_LEN];
1375 	u_char addrhi[ETHER_ADDR_LEN];
1376 	int error;
1377 
1378 	ETHER_LOCK(ec);
1379 	error = ether_multiaddr(sa, addrlo, addrhi);
1380 	if (error != 0)
1381 		goto error;
1382 
1383 	/*
1384 	 * Look up the address in our list.
1385 	 */
1386 	enm = ether_lookup_multi(addrlo, addrhi, ec);
1387 	if (enm == NULL) {
1388 		error = ENXIO;
1389 		goto error;
1390 	}
1391 	if (--enm->enm_refcount != 0) {
1392 		/*
1393 		 * Still some claims to this record.
1394 		 */
1395 		error = 0;
1396 		goto error;
1397 	}
1398 
1399 	/*
1400 	 * No remaining claims to this record; unlink and free it.
1401 	 */
1402 	LIST_REMOVE(enm, enm_list);
1403 	ec->ec_multicnt--;
1404 	ETHER_UNLOCK(ec);
1405 	kmem_free(enm, sizeof(*enm));
1406 
1407 	/*
1408 	 * Return ENETRESET to inform the driver that the list has changed
1409 	 * and its reception filter should be adjusted accordingly.
1410 	 */
1411 	return ENETRESET;
1412 
1413 error:
1414 	ETHER_UNLOCK(ec);
1415 	return error;
1416 }
1417 
1418 void
1419 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb)
1420 {
1421 	ec->ec_ifflags_cb = cb;
1422 }
1423 
1424 void
1425 ether_set_vlan_cb(struct ethercom *ec, ether_vlancb_t cb)
1426 {
1427 
1428 	ec->ec_vlan_cb = cb;
1429 }
1430 
1431 static int
1432 ether_ioctl_reinit(struct ethercom *ec)
1433 {
1434 	struct ifnet *ifp = &ec->ec_if;
1435 	int error;
1436 
1437 	KASSERTMSG(IFNET_LOCKED(ifp), "%s", ifp->if_xname);
1438 
1439 	switch (ifp->if_flags & (IFF_UP | IFF_RUNNING)) {
1440 	case IFF_RUNNING:
1441 		/*
1442 		 * If interface is marked down and it is running,
1443 		 * then stop and disable it.
1444 		 */
1445 		if_stop(ifp, 1);
1446 		break;
1447 	case IFF_UP:
1448 		/*
1449 		 * If interface is marked up and it is stopped, then
1450 		 * start it.
1451 		 */
1452 		return if_init(ifp);
1453 	case IFF_UP | IFF_RUNNING:
1454 		error = 0;
1455 		if (ec->ec_ifflags_cb != NULL) {
1456 			error = (*ec->ec_ifflags_cb)(ec);
1457 			if (error == ENETRESET) {
1458 				/*
1459 				 * Reset the interface to pick up
1460 				 * changes in any other flags that
1461 				 * affect the hardware state.
1462 				 */
1463 				return if_init(ifp);
1464 			}
1465 		} else
1466 			error = if_init(ifp);
1467 		return error;
1468 	case 0:
1469 		break;
1470 	}
1471 
1472 	return 0;
1473 }
1474 
1475 /*
1476  * Common ioctls for Ethernet interfaces.  Note, we must be
1477  * called at splnet().
1478  */
1479 int
1480 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data)
1481 {
1482 	struct ethercom *ec = (void *)ifp;
1483 	struct eccapreq *eccr;
1484 	struct ifreq *ifr = (struct ifreq *)data;
1485 	struct if_laddrreq *iflr = data;
1486 	const struct sockaddr_dl *sdl;
1487 	static const uint8_t zero[ETHER_ADDR_LEN];
1488 	int error;
1489 
1490 	switch (cmd) {
1491 	case SIOCINITIFADDR:
1492 	    {
1493 		struct ifaddr *ifa = (struct ifaddr *)data;
1494 		if (ifa->ifa_addr->sa_family != AF_LINK
1495 		    && (ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
1496 		       (IFF_UP | IFF_RUNNING)) {
1497 			ifp->if_flags |= IFF_UP;
1498 			if ((error = if_init(ifp)) != 0)
1499 				return error;
1500 		}
1501 #ifdef INET
1502 		if (ifa->ifa_addr->sa_family == AF_INET)
1503 			arp_ifinit(ifp, ifa);
1504 #endif
1505 		return 0;
1506 	    }
1507 
1508 	case SIOCSIFMTU:
1509 	    {
1510 		int maxmtu;
1511 
1512 		if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU)
1513 			maxmtu = ETHERMTU_JUMBO;
1514 		else
1515 			maxmtu = ETHERMTU;
1516 
1517 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu)
1518 			return EINVAL;
1519 		else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET)
1520 			return error;
1521 		else if (ifp->if_flags & IFF_UP) {
1522 			/* Make sure the device notices the MTU change. */
1523 			return if_init(ifp);
1524 		} else
1525 			return 0;
1526 	    }
1527 
1528 	case SIOCSIFFLAGS:
1529 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
1530 			return error;
1531 		return ether_ioctl_reinit(ec);
1532 	case SIOCGIFFLAGS:
1533 		error = ifioctl_common(ifp, cmd, data);
1534 		if (error == 0) {
1535 			/* Set IFF_ALLMULTI for backcompat */
1536 			ifr->ifr_flags |= (ec->ec_flags & ETHER_F_ALLMULTI) ?
1537 			    IFF_ALLMULTI : 0;
1538 		}
1539 		return error;
1540 	case SIOCGETHERCAP:
1541 		eccr = (struct eccapreq *)data;
1542 		eccr->eccr_capabilities = ec->ec_capabilities;
1543 		eccr->eccr_capenable = ec->ec_capenable;
1544 		return 0;
1545 	case SIOCSETHERCAP:
1546 		eccr = (struct eccapreq *)data;
1547 		if ((eccr->eccr_capenable & ~ec->ec_capabilities) != 0)
1548 			return EINVAL;
1549 		if (eccr->eccr_capenable == ec->ec_capenable)
1550 			return 0;
1551 #if 0 /* notyet */
1552 		ec->ec_capenable = (ec->ec_capenable & ETHERCAP_CANTCHANGE)
1553 		    | (eccr->eccr_capenable & ~ETHERCAP_CANTCHANGE);
1554 #else
1555 		ec->ec_capenable = eccr->eccr_capenable;
1556 #endif
1557 		return ether_ioctl_reinit(ec);
1558 	case SIOCADDMULTI:
1559 		return ether_addmulti(ifreq_getaddr(cmd, ifr), ec);
1560 	case SIOCDELMULTI:
1561 		return ether_delmulti(ifreq_getaddr(cmd, ifr), ec);
1562 	case SIOCSIFMEDIA:
1563 	case SIOCGIFMEDIA:
1564 		if (ec->ec_mii != NULL)
1565 			return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media,
1566 			    cmd);
1567 		else if (ec->ec_ifmedia != NULL)
1568 			return ifmedia_ioctl(ifp, ifr, ec->ec_ifmedia, cmd);
1569 		else
1570 			return ENOTTY;
1571 		break;
1572 	case SIOCALIFADDR:
1573 		sdl = satocsdl(sstocsa(&iflr->addr));
1574 		if (sdl->sdl_family != AF_LINK)
1575 			;
1576 		else if (ETHER_IS_MULTICAST(CLLADDR(sdl)))
1577 			return EINVAL;
1578 		else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0)
1579 			return EINVAL;
1580 		/*FALLTHROUGH*/
1581 	default:
1582 		return ifioctl_common(ifp, cmd, data);
1583 	}
1584 	return 0;
1585 }
1586 
1587 /*
1588  * Enable/disable passing VLAN packets if the parent interface supports it.
1589  * Return:
1590  * 	 0: Ok
1591  *	-1: Parent interface does not support vlans
1592  *	>0: Error
1593  */
1594 int
1595 ether_enable_vlan_mtu(struct ifnet *ifp)
1596 {
1597 	int error;
1598 	struct ethercom *ec = (void *)ifp;
1599 
1600 	/* Parent does not support VLAN's */
1601 	if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0)
1602 		return -1;
1603 
1604 	/*
1605 	 * Parent supports the VLAN_MTU capability,
1606 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames;
1607 	 * enable it.
1608 	 */
1609 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1610 
1611 	/* Interface is down, defer for later */
1612 	if ((ifp->if_flags & IFF_UP) == 0)
1613 		return 0;
1614 
1615 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1616 		return 0;
1617 
1618 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1619 	return error;
1620 }
1621 
1622 int
1623 ether_disable_vlan_mtu(struct ifnet *ifp)
1624 {
1625 	int error;
1626 	struct ethercom *ec = (void *)ifp;
1627 
1628 	/* We still have VLAN's, defer for later */
1629 	if (ec->ec_nvlans != 0)
1630 		return 0;
1631 
1632 	/* Parent does not support VLAB's, nothing to do. */
1633 	if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0)
1634 		return -1;
1635 
1636 	/*
1637 	 * Disable Tx/Rx of VLAN-sized frames.
1638 	 */
1639 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1640 
1641 	/* Interface is down, defer for later */
1642 	if ((ifp->if_flags & IFF_UP) == 0)
1643 		return 0;
1644 
1645 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1646 		return 0;
1647 
1648 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1649 	return error;
1650 }
1651 
1652 /*
1653  * Add and delete VLAN TAG
1654  */
1655 int
1656 ether_add_vlantag(struct ifnet *ifp, uint16_t vtag, bool *vlanmtu_status)
1657 {
1658 	struct ethercom *ec = (void *)ifp;
1659 	struct vlanid_list *vidp;
1660 	bool vlanmtu_enabled;
1661 	uint16_t vid = EVL_VLANOFTAG(vtag);
1662 	int error;
1663 
1664 	vlanmtu_enabled = false;
1665 
1666 	/* Add a vid to the list */
1667 	vidp = kmem_alloc(sizeof(*vidp), KM_SLEEP);
1668 	vidp->vid = vid;
1669 
1670 	ETHER_LOCK(ec);
1671 	ec->ec_nvlans++;
1672 	SIMPLEQ_INSERT_TAIL(&ec->ec_vids, vidp, vid_list);
1673 	ETHER_UNLOCK(ec);
1674 
1675 	if (ec->ec_nvlans == 1) {
1676 		IFNET_LOCK(ifp);
1677 		error = ether_enable_vlan_mtu(ifp);
1678 		IFNET_UNLOCK(ifp);
1679 
1680 		if (error == 0) {
1681 			vlanmtu_enabled = true;
1682 		} else if (error != -1) {
1683 			goto fail;
1684 		}
1685 	}
1686 
1687 	if (ec->ec_vlan_cb != NULL) {
1688 		error = (*ec->ec_vlan_cb)(ec, vid, true);
1689 		if (error != 0)
1690 			goto fail;
1691 	}
1692 
1693 	if (vlanmtu_status != NULL)
1694 		*vlanmtu_status = vlanmtu_enabled;
1695 
1696 	return 0;
1697 fail:
1698 	ETHER_LOCK(ec);
1699 	ec->ec_nvlans--;
1700 	SIMPLEQ_REMOVE(&ec->ec_vids, vidp, vlanid_list, vid_list);
1701 	ETHER_UNLOCK(ec);
1702 
1703 	if (vlanmtu_enabled) {
1704 		IFNET_LOCK(ifp);
1705 		(void)ether_disable_vlan_mtu(ifp);
1706 		IFNET_UNLOCK(ifp);
1707 	}
1708 
1709 	kmem_free(vidp, sizeof(*vidp));
1710 
1711 	return error;
1712 }
1713 
1714 int
1715 ether_del_vlantag(struct ifnet *ifp, uint16_t vtag)
1716 {
1717 	struct ethercom *ec = (void *)ifp;
1718 	struct vlanid_list *vidp;
1719 	uint16_t vid = EVL_VLANOFTAG(vtag);
1720 
1721 	ETHER_LOCK(ec);
1722 	SIMPLEQ_FOREACH(vidp, &ec->ec_vids, vid_list) {
1723 		if (vidp->vid == vid) {
1724 			SIMPLEQ_REMOVE(&ec->ec_vids, vidp,
1725 			    vlanid_list, vid_list);
1726 			ec->ec_nvlans--;
1727 			break;
1728 		}
1729 	}
1730 	ETHER_UNLOCK(ec);
1731 
1732 	if (vidp == NULL)
1733 		return ENOENT;
1734 
1735 	if (ec->ec_vlan_cb != NULL) {
1736 		(void)(*ec->ec_vlan_cb)(ec, vidp->vid, false);
1737 	}
1738 
1739 	if (ec->ec_nvlans == 0) {
1740 		IFNET_LOCK(ifp);
1741 		(void)ether_disable_vlan_mtu(ifp);
1742 		IFNET_UNLOCK(ifp);
1743 	}
1744 
1745 	kmem_free(vidp, sizeof(*vidp));
1746 
1747 	return 0;
1748 }
1749 
1750 static int
1751 ether_multicast_sysctl(SYSCTLFN_ARGS)
1752 {
1753 	struct ether_multi *enm;
1754 	struct ifnet *ifp;
1755 	struct ethercom *ec;
1756 	int error = 0;
1757 	size_t written;
1758 	struct psref psref;
1759 	int bound;
1760 	unsigned int multicnt;
1761 	struct ether_multi_sysctl *addrs;
1762 	int i;
1763 
1764 	if (namelen != 1)
1765 		return EINVAL;
1766 
1767 	bound = curlwp_bind();
1768 	ifp = if_get_byindex(name[0], &psref);
1769 	if (ifp == NULL) {
1770 		error = ENODEV;
1771 		goto out;
1772 	}
1773 	if (ifp->if_type != IFT_ETHER) {
1774 		if_put(ifp, &psref);
1775 		*oldlenp = 0;
1776 		goto out;
1777 	}
1778 	ec = (struct ethercom *)ifp;
1779 
1780 	if (oldp == NULL) {
1781 		if_put(ifp, &psref);
1782 		*oldlenp = ec->ec_multicnt * sizeof(*addrs);
1783 		goto out;
1784 	}
1785 
1786 	/*
1787 	 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which
1788 	 * is sleepable, while holding it. Copy data to a local buffer first
1789 	 * with the lock taken and then call sysctl_copyout without holding it.
1790 	 */
1791 retry:
1792 	multicnt = ec->ec_multicnt;
1793 
1794 	if (multicnt == 0) {
1795 		if_put(ifp, &psref);
1796 		*oldlenp = 0;
1797 		goto out;
1798 	}
1799 
1800 	addrs = kmem_zalloc(sizeof(*addrs) * multicnt, KM_SLEEP);
1801 
1802 	ETHER_LOCK(ec);
1803 	if (multicnt != ec->ec_multicnt) {
1804 		/* The number of multicast addresses has changed */
1805 		ETHER_UNLOCK(ec);
1806 		kmem_free(addrs, sizeof(*addrs) * multicnt);
1807 		goto retry;
1808 	}
1809 
1810 	i = 0;
1811 	LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) {
1812 		struct ether_multi_sysctl *addr = &addrs[i];
1813 		addr->enm_refcount = enm->enm_refcount;
1814 		memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN);
1815 		memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN);
1816 		i++;
1817 	}
1818 	ETHER_UNLOCK(ec);
1819 
1820 	error = 0;
1821 	written = 0;
1822 	for (i = 0; i < multicnt; i++) {
1823 		struct ether_multi_sysctl *addr = &addrs[i];
1824 
1825 		if (written + sizeof(*addr) > *oldlenp)
1826 			break;
1827 		error = sysctl_copyout(l, addr, oldp, sizeof(*addr));
1828 		if (error)
1829 			break;
1830 		written += sizeof(*addr);
1831 		oldp = (char *)oldp + sizeof(*addr);
1832 	}
1833 	kmem_free(addrs, sizeof(*addrs) * multicnt);
1834 
1835 	if_put(ifp, &psref);
1836 
1837 	*oldlenp = written;
1838 out:
1839 	curlwp_bindx(bound);
1840 	return error;
1841 }
1842 
1843 static void
1844 ether_sysctl_setup(struct sysctllog **clog)
1845 {
1846 	const struct sysctlnode *rnode = NULL;
1847 
1848 	sysctl_createv(clog, 0, NULL, &rnode,
1849 		       CTLFLAG_PERMANENT,
1850 		       CTLTYPE_NODE, "ether",
1851 		       SYSCTL_DESCR("Ethernet-specific information"),
1852 		       NULL, 0, NULL, 0,
1853 		       CTL_NET, CTL_CREATE, CTL_EOL);
1854 
1855 	sysctl_createv(clog, 0, &rnode, NULL,
1856 		       CTLFLAG_PERMANENT,
1857 		       CTLTYPE_NODE, "multicast",
1858 		       SYSCTL_DESCR("multicast addresses"),
1859 		       ether_multicast_sysctl, 0, NULL, 0,
1860 		       CTL_CREATE, CTL_EOL);
1861 
1862 	sysctl_createv(clog, 0, &rnode, NULL,
1863 		       CTLFLAG_PERMANENT | CTLFLAG_READWRITE,
1864 		       CTLTYPE_STRING, "rps_hash",
1865 		       SYSCTL_DESCR("Interface rps hash function control"),
1866 		       sysctl_pktq_rps_hash_handler, 0, (void *)&ether_pktq_rps_hash_p,
1867 		       PKTQ_RPS_HASH_NAME_LEN,
1868 		       CTL_CREATE, CTL_EOL);
1869 }
1870 
1871 void
1872 etherinit(void)
1873 {
1874 
1875 #ifdef DIAGNOSTIC
1876 	mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET);
1877 #endif
1878 	ether_pktq_rps_hash_p = pktq_rps_hash_default;
1879 	ether_sysctl_setup(NULL);
1880 }
1881