xref: /netbsd-src/sys/net/if_ethersubr.c (revision be7d6d28fc2bc4242e902b8215a446332243c607)
1 /*	$NetBSD: if_ethersubr.c,v 1.276 2019/07/17 03:26:24 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 1.276 2019/07/17 03:26:24 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/rnd.h>
89 #include <sys/rndsource.h>
90 #include <sys/cpu.h>
91 #include <sys/kmem.h>
92 
93 #include <net/if.h>
94 #include <net/netisr.h>
95 #include <net/route.h>
96 #include <net/if_llc.h>
97 #include <net/if_dl.h>
98 #include <net/if_types.h>
99 #include <net/pktqueue.h>
100 
101 #include <net/if_media.h>
102 #include <dev/mii/mii.h>
103 #include <dev/mii/miivar.h>
104 
105 #if NARP == 0
106 /*
107  * XXX there should really be a way to issue this warning from within config(8)
108  */
109 #error You have included NETATALK or a pseudo-device in your configuration that depends on the presence of ethernet interfaces, but have no such interfaces configured. Check if you really need pseudo-device bridge, pppoe, vlan or options NETATALK.
110 #endif
111 
112 #include <net/bpf.h>
113 
114 #include <net/if_ether.h>
115 #include <net/if_vlanvar.h>
116 
117 #if NPPPOE > 0
118 #include <net/if_pppoe.h>
119 #endif
120 
121 #if NAGR > 0
122 #include <net/agr/ieee8023_slowprotocols.h>	/* XXX */
123 #include <net/agr/ieee8023ad.h>
124 #include <net/agr/if_agrvar.h>
125 #endif
126 
127 #if NBRIDGE > 0
128 #include <net/if_bridgevar.h>
129 #endif
130 
131 #include <netinet/in.h>
132 #ifdef INET
133 #include <netinet/in_var.h>
134 #endif
135 #include <netinet/if_inarp.h>
136 
137 #ifdef INET6
138 #ifndef INET
139 #include <netinet/in.h>
140 #endif
141 #include <netinet6/in6_var.h>
142 #include <netinet6/nd6.h>
143 #endif
144 
145 #include "carp.h"
146 #if NCARP > 0
147 #include <netinet/ip_carp.h>
148 #endif
149 
150 #ifdef NETATALK
151 #include <netatalk/at.h>
152 #include <netatalk/at_var.h>
153 #include <netatalk/at_extern.h>
154 
155 #define llc_snap_org_code llc_un.type_snap.org_code
156 #define llc_snap_ether_type llc_un.type_snap.ether_type
157 
158 extern u_char	at_org_code[3];
159 extern u_char	aarp_org_code[3];
160 #endif /* NETATALK */
161 
162 #ifdef MPLS
163 #include <netmpls/mpls.h>
164 #include <netmpls/mpls_var.h>
165 #endif
166 
167 static struct timeval bigpktppslim_last;
168 static int bigpktppslim = 2;	/* XXX */
169 static int bigpktpps_count;
170 static kmutex_t bigpktpps_lock __cacheline_aligned;
171 
172 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] =
173     { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
174 const uint8_t ethermulticastaddr_slowprotocols[ETHER_ADDR_LEN] =
175     { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x02 };
176 #define senderr(e) { error = (e); goto bad;}
177 
178 static int ether_output(struct ifnet *, struct mbuf *,
179     const struct sockaddr *, const struct rtentry *);
180 
181 /*
182  * Ethernet output routine.
183  * Encapsulate a packet of type family for the local net.
184  * Assumes that ifp is actually pointer to ethercom structure.
185  */
186 static int
187 ether_output(struct ifnet * const ifp0, struct mbuf * const m0,
188     const struct sockaddr * const dst, const struct rtentry *rt)
189 {
190 	uint8_t esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
191 	uint16_t etype = 0;
192 	int error = 0, hdrcmplt = 0;
193 	struct mbuf *m = m0;
194 	struct mbuf *mcopy = NULL;
195 	struct ether_header *eh;
196 	struct ifnet *ifp = ifp0;
197 #ifdef INET
198 	struct arphdr *ah;
199 #endif
200 #ifdef NETATALK
201 	struct at_ifaddr *aa;
202 #endif
203 
204 #ifdef MBUFTRACE
205 	m_claimm(m, ifp->if_mowner);
206 #endif
207 
208 #if NCARP > 0
209 	if (ifp->if_type == IFT_CARP) {
210 		struct ifaddr *ifa;
211 		int s = pserialize_read_enter();
212 
213 		/* loop back if this is going to the carp interface */
214 		if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP &&
215 		    (ifa = ifa_ifwithaddr(dst)) != NULL) {
216 			if (ifa->ifa_ifp == ifp0) {
217 				pserialize_read_exit(s);
218 				return looutput(ifp0, m, dst, rt);
219 			}
220 		}
221 		pserialize_read_exit(s);
222 
223 		ifp = ifp->if_carpdev;
224 		/* ac = (struct arpcom *)ifp; */
225 
226 		if ((ifp0->if_flags & (IFF_UP | IFF_RUNNING)) !=
227 		    (IFF_UP | IFF_RUNNING))
228 			senderr(ENETDOWN);
229 	}
230 #endif
231 
232 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
233 		senderr(ENETDOWN);
234 
235 	switch (dst->sa_family) {
236 
237 #ifdef INET
238 	case AF_INET:
239 		if (m->m_flags & M_BCAST) {
240 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
241 		} else if (m->m_flags & M_MCAST) {
242 			ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst);
243 		} else {
244 			error = arpresolve(ifp, rt, m, dst, edst, sizeof(edst));
245 			if (error)
246 				return (error == EWOULDBLOCK) ? 0 : error;
247 		}
248 		/* If broadcasting on a simplex interface, loopback a copy */
249 		if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
250 			mcopy = m_copypacket(m, M_DONTWAIT);
251 		etype = htons(ETHERTYPE_IP);
252 		break;
253 
254 	case AF_ARP:
255 		ah = mtod(m, struct arphdr *);
256 		if (m->m_flags & M_BCAST) {
257 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
258 		} else {
259 			void *tha = ar_tha(ah);
260 
261 			if (tha == NULL) {
262 				/* fake with ARPHRD_IEEE1394 */
263 				m_freem(m);
264 				return 0;
265 			}
266 			memcpy(edst, tha, sizeof(edst));
267 		}
268 
269 		ah->ar_hrd = htons(ARPHRD_ETHER);
270 
271 		switch (ntohs(ah->ar_op)) {
272 		case ARPOP_REVREQUEST:
273 		case ARPOP_REVREPLY:
274 			etype = htons(ETHERTYPE_REVARP);
275 			break;
276 
277 		case ARPOP_REQUEST:
278 		case ARPOP_REPLY:
279 		default:
280 			etype = htons(ETHERTYPE_ARP);
281 		}
282 		break;
283 #endif
284 
285 #ifdef INET6
286 	case AF_INET6:
287 		if (m->m_flags & M_BCAST) {
288 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
289 		} else if (m->m_flags & M_MCAST) {
290 			ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr,
291 			    edst);
292 		} else {
293 			error = nd6_resolve(ifp, rt, m, dst, edst,
294 			    sizeof(edst));
295 			if (error)
296 				return (error == EWOULDBLOCK) ? 0 : error;
297 		}
298 		etype = htons(ETHERTYPE_IPV6);
299 		break;
300 #endif
301 
302 #ifdef NETATALK
303 	case AF_APPLETALK: {
304 		struct ifaddr *ifa;
305 		int s;
306 
307 		KERNEL_LOCK(1, NULL);
308 
309 		if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) {
310 			KERNEL_UNLOCK_ONE(NULL);
311 			return 0;
312 		}
313 
314 		/*
315 		 * ifaddr is the first thing in at_ifaddr
316 		 */
317 		s = pserialize_read_enter();
318 		ifa = at_ifawithnet((const struct sockaddr_at *)dst, ifp);
319 		if (ifa == NULL) {
320 			pserialize_read_exit(s);
321 			KERNEL_UNLOCK_ONE(NULL);
322 			senderr(EADDRNOTAVAIL);
323 		}
324 		aa = (struct at_ifaddr *)ifa;
325 
326 		/*
327 		 * In the phase 2 case, we need to prepend an mbuf for the
328 		 * llc header.
329 		 */
330 		if (aa->aa_flags & AFA_PHASE2) {
331 			struct llc llc;
332 
333 			M_PREPEND(m, sizeof(struct llc), M_DONTWAIT);
334 			if (m == NULL) {
335 				pserialize_read_exit(s);
336 				KERNEL_UNLOCK_ONE(NULL);
337 				senderr(ENOBUFS);
338 			}
339 
340 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
341 			llc.llc_control = LLC_UI;
342 			memcpy(llc.llc_snap_org_code, at_org_code,
343 			    sizeof(llc.llc_snap_org_code));
344 			llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK);
345 			memcpy(mtod(m, void *), &llc, sizeof(struct llc));
346 		} else {
347 			etype = htons(ETHERTYPE_ATALK);
348 		}
349 		pserialize_read_exit(s);
350 		KERNEL_UNLOCK_ONE(NULL);
351 		break;
352 	}
353 #endif /* NETATALK */
354 
355 	case pseudo_AF_HDRCMPLT:
356 		hdrcmplt = 1;
357 		memcpy(esrc,
358 		    ((const struct ether_header *)dst->sa_data)->ether_shost,
359 		    sizeof(esrc));
360 		/* FALLTHROUGH */
361 
362 	case AF_UNSPEC:
363 		memcpy(edst,
364 		    ((const struct ether_header *)dst->sa_data)->ether_dhost,
365 		    sizeof(edst));
366 		/* AF_UNSPEC doesn't swap the byte order of the ether_type. */
367 		etype = ((const struct ether_header *)dst->sa_data)->ether_type;
368 		break;
369 
370 	default:
371 		printf("%s: can't handle af%d\n", ifp->if_xname,
372 		    dst->sa_family);
373 		senderr(EAFNOSUPPORT);
374 	}
375 
376 #ifdef MPLS
377 	{
378 		struct m_tag *mtag;
379 		mtag = m_tag_find(m, PACKET_TAG_MPLS);
380 		if (mtag != NULL) {
381 			/* Having the tag itself indicates it's MPLS */
382 			etype = htons(ETHERTYPE_MPLS);
383 			m_tag_delete(m, mtag);
384 		}
385 	}
386 #endif
387 
388 	if (mcopy)
389 		(void)looutput(ifp, mcopy, dst, rt);
390 
391 	KASSERT((m->m_flags & M_PKTHDR) != 0);
392 
393 	/*
394 	 * If no ether type is set, this must be a 802.2 formatted packet.
395 	 */
396 	if (etype == 0)
397 		etype = htons(m->m_pkthdr.len);
398 
399 	/*
400 	 * Add local net header. If no space in first mbuf, allocate another.
401 	 */
402 	M_PREPEND(m, sizeof(struct ether_header), M_DONTWAIT);
403 	if (m == NULL)
404 		senderr(ENOBUFS);
405 
406 	eh = mtod(m, struct ether_header *);
407 	/* Note: etype is already in network byte order. */
408 	memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
409 	memcpy(eh->ether_dhost, edst, sizeof(edst));
410 	if (hdrcmplt) {
411 		memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost));
412 	} else {
413 	 	memcpy(eh->ether_shost, CLLADDR(ifp->if_sadl),
414 		    sizeof(eh->ether_shost));
415 	}
416 
417 #if NCARP > 0
418 	if (ifp0 != ifp && ifp0->if_type == IFT_CARP) {
419 	 	memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl),
420 		    sizeof(eh->ether_shost));
421 	}
422 #endif
423 
424 	if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0)
425 		return error;
426 	if (m == NULL)
427 		return 0;
428 
429 #if NBRIDGE > 0
430 	/*
431 	 * Bridges require special output handling.
432 	 */
433 	if (ifp->if_bridge)
434 		return bridge_output(ifp, m, NULL, NULL);
435 #endif
436 
437 #if NCARP > 0
438 	if (ifp != ifp0)
439 		ifp0->if_obytes += m->m_pkthdr.len + ETHER_HDR_LEN;
440 #endif
441 
442 #ifdef ALTQ
443 	KERNEL_LOCK(1, NULL);
444 	/*
445 	 * If ALTQ is enabled on the parent interface, do
446 	 * classification; the queueing discipline might not
447 	 * require classification, but might require the
448 	 * address family/header pointer in the pktattr.
449 	 */
450 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
451 		altq_etherclassify(&ifp->if_snd, m);
452 	KERNEL_UNLOCK_ONE(NULL);
453 #endif
454 	return ifq_enqueue(ifp, m);
455 
456 bad:
457 	if (m)
458 		m_freem(m);
459 	return error;
460 }
461 
462 #ifdef ALTQ
463 /*
464  * This routine is a slight hack to allow a packet to be classified
465  * if the Ethernet headers are present.  It will go away when ALTQ's
466  * classification engine understands link headers.
467  *
468  * XXX: We may need to do m_pullups here. First to ensure struct ether_header
469  * is indeed contiguous, then to read the LLC and so on.
470  */
471 void
472 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m)
473 {
474 	struct ether_header *eh;
475 	struct mbuf *mtop = m;
476 	uint16_t ether_type;
477 	int hlen, af, hdrsize;
478 	void *hdr;
479 
480 	KASSERT((mtop->m_flags & M_PKTHDR) != 0);
481 
482 	hlen = ETHER_HDR_LEN;
483 	eh = mtod(m, struct ether_header *);
484 
485 	ether_type = htons(eh->ether_type);
486 
487 	if (ether_type < ETHERMTU) {
488 		/* LLC/SNAP */
489 		struct llc *llc = (struct llc *)(eh + 1);
490 		hlen += 8;
491 
492 		if (m->m_len < hlen ||
493 		    llc->llc_dsap != LLC_SNAP_LSAP ||
494 		    llc->llc_ssap != LLC_SNAP_LSAP ||
495 		    llc->llc_control != LLC_UI) {
496 			/* Not SNAP. */
497 			goto bad;
498 		}
499 
500 		ether_type = htons(llc->llc_un.type_snap.ether_type);
501 	}
502 
503 	switch (ether_type) {
504 	case ETHERTYPE_IP:
505 		af = AF_INET;
506 		hdrsize = 20;		/* sizeof(struct ip) */
507 		break;
508 
509 	case ETHERTYPE_IPV6:
510 		af = AF_INET6;
511 		hdrsize = 40;		/* sizeof(struct ip6_hdr) */
512 		break;
513 
514 	default:
515 		af = AF_UNSPEC;
516 		hdrsize = 0;
517 		break;
518 	}
519 
520 	while (m->m_len <= hlen) {
521 		hlen -= m->m_len;
522 		m = m->m_next;
523 		if (m == NULL)
524 			goto bad;
525 	}
526 
527 	if (m->m_len < (hlen + hdrsize)) {
528 		/*
529 		 * protocol header not in a single mbuf.
530 		 * We can't cope with this situation right
531 		 * now (but it shouldn't ever happen, really, anyhow).
532 		 */
533 #ifdef DEBUG
534 		printf("altq_etherclassify: headers span multiple mbufs: "
535 		    "%d < %d\n", m->m_len, (hlen + hdrsize));
536 #endif
537 		goto bad;
538 	}
539 
540 	m->m_data += hlen;
541 	m->m_len -= hlen;
542 
543 	hdr = mtod(m, void *);
544 
545 	if (ALTQ_NEEDS_CLASSIFY(ifq)) {
546 		mtop->m_pkthdr.pattr_class =
547 		    (*ifq->altq_classify)(ifq->altq_clfier, m, af);
548 	}
549 	mtop->m_pkthdr.pattr_af = af;
550 	mtop->m_pkthdr.pattr_hdr = hdr;
551 
552 	m->m_data -= hlen;
553 	m->m_len += hlen;
554 
555 	return;
556 
557 bad:
558 	mtop->m_pkthdr.pattr_class = NULL;
559 	mtop->m_pkthdr.pattr_hdr = NULL;
560 	mtop->m_pkthdr.pattr_af = AF_UNSPEC;
561 }
562 #endif /* ALTQ */
563 
564 /*
565  * Process a received Ethernet packet;
566  * the packet is in the mbuf chain m with
567  * the ether header.
568  */
569 void
570 ether_input(struct ifnet *ifp, struct mbuf *m)
571 {
572 	struct ethercom *ec = (struct ethercom *) ifp;
573 	pktqueue_t *pktq = NULL;
574 	struct ifqueue *inq = NULL;
575 	uint16_t etype;
576 	struct ether_header *eh;
577 	size_t ehlen;
578 	static int earlypkts;
579 	int isr = 0;
580 #if defined (LLC) || defined(NETATALK)
581 	struct llc *l;
582 #endif
583 
584 	KASSERT(!cpu_intr_p());
585 	KASSERT((m->m_flags & M_PKTHDR) != 0);
586 
587 	if ((ifp->if_flags & IFF_UP) == 0) {
588 		m_freem(m);
589 		return;
590 	}
591 	if (m->m_len < sizeof(*eh)) {
592 		m = m_pullup(m, sizeof(*eh));
593 		if (m == NULL)
594 			return;
595 	}
596 
597 #ifdef MBUFTRACE
598 	m_claimm(m, &ec->ec_rx_mowner);
599 #endif
600 	eh = mtod(m, struct ether_header *);
601 	etype = ntohs(eh->ether_type);
602 	ehlen = sizeof(*eh);
603 
604 	if (__predict_false(earlypkts < 100 || !rnd_initial_entropy)) {
605 		rnd_add_data(NULL, eh, ehlen, 0);
606 		earlypkts++;
607 	}
608 
609 	/*
610 	 * Determine if the packet is within its size limits. For MPLS the
611 	 * header length is variable, so we skip the check.
612 	 */
613 	if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len >
614 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
615 		mutex_enter(&bigpktpps_lock);
616 		if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count,
617 		    bigpktppslim)) {
618 			printf("%s: discarding oversize frame (len=%d)\n",
619 			    ifp->if_xname, m->m_pkthdr.len);
620 		}
621 		mutex_exit(&bigpktpps_lock);
622 		m_freem(m);
623 		return;
624 	}
625 
626 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
627 		/*
628 		 * If this is not a simplex interface, drop the packet
629 		 * if it came from us.
630 		 */
631 		if ((ifp->if_flags & IFF_SIMPLEX) == 0 &&
632 		    memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost,
633 		    ETHER_ADDR_LEN) == 0) {
634 			m_freem(m);
635 			return;
636 		}
637 
638 		if (memcmp(etherbroadcastaddr,
639 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
640 			m->m_flags |= M_BCAST;
641 		else
642 			m->m_flags |= M_MCAST;
643 		ifp->if_imcasts++;
644 	}
645 
646 	/* If the CRC is still on the packet, trim it off. */
647 	if (m->m_flags & M_HASFCS) {
648 		m_adj(m, -ETHER_CRC_LEN);
649 		m->m_flags &= ~M_HASFCS;
650 	}
651 
652 	ifp->if_ibytes += m->m_pkthdr.len;
653 
654 #if NCARP > 0
655 	if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) {
656 		/*
657 		 * Clear M_PROMISC, in case the packet comes from a
658 		 * vlan.
659 		 */
660 		m->m_flags &= ~M_PROMISC;
661 		if (carp_input(m, (uint8_t *)&eh->ether_shost,
662 		    (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0)
663 			return;
664 	}
665 #endif
666 
667 	if ((m->m_flags & (M_BCAST | M_MCAST | M_PROMISC)) == 0 &&
668 	    (ifp->if_flags & IFF_PROMISC) != 0 &&
669 	    memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost,
670 	     ETHER_ADDR_LEN) != 0) {
671 		m->m_flags |= M_PROMISC;
672 	}
673 
674 	if ((m->m_flags & M_PROMISC) == 0) {
675 		if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
676 			return;
677 		if (m == NULL)
678 			return;
679 
680 		eh = mtod(m, struct ether_header *);
681 		etype = ntohs(eh->ether_type);
682 	}
683 
684 #if NAGR > 0
685 	if (ifp->if_agrprivate &&
686 	    __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) {
687 		m->m_flags &= ~M_PROMISC;
688 		agr_input(ifp, m);
689 		return;
690 	}
691 #endif
692 
693 	/*
694 	 * If VLANs are configured on the interface, check to
695 	 * see if the device performed the decapsulation and
696 	 * provided us with the tag.
697 	 */
698 	if (ec->ec_nvlans && vlan_has_tag(m)) {
699 #if NVLAN > 0
700 		/*
701 		 * vlan_input() will either recursively call ether_input()
702 		 * or drop the packet.
703 		 */
704 		vlan_input(ifp, m);
705 #else
706 		m_freem(m);
707 #endif
708 		return;
709 	}
710 
711 	/*
712 	 * Handle protocols that expect to have the Ethernet header
713 	 * (and possibly FCS) intact.
714 	 */
715 	switch (etype) {
716 	case ETHERTYPE_VLAN: {
717 		struct ether_vlan_header *evl = (void *)eh;
718 
719 		/*
720 		 * If there is a tag of 0, then the VLAN header was probably
721 		 * just being used to store the priority.  Extract the ether
722 		 * type, and if IP or IPV6, let them deal with it.
723 		 */
724 		if (m->m_len >= sizeof(*evl) &&
725 		    EVL_VLANOFTAG(evl->evl_tag) == 0) {
726 			etype = ntohs(evl->evl_proto);
727 			ehlen = sizeof(*evl);
728 			if ((m->m_flags & M_PROMISC) == 0 &&
729 			    (etype == ETHERTYPE_IP ||
730 			     etype == ETHERTYPE_IPV6))
731 				break;
732 		}
733 
734 #if NVLAN > 0
735 		/*
736 		 * vlan_input() will either recursively call ether_input()
737 		 * or drop the packet.
738 		 */
739 		if (ec->ec_nvlans != 0)
740 			vlan_input(ifp, m);
741 		else
742 #endif
743 			m_freem(m);
744 
745 		return;
746 	}
747 
748 #if NPPPOE > 0
749 	case ETHERTYPE_PPPOEDISC:
750 		pppoedisc_input(ifp, m);
751 		return;
752 
753 	case ETHERTYPE_PPPOE:
754 		pppoe_input(ifp, m);
755 		return;
756 #endif
757 
758 	case ETHERTYPE_SLOWPROTOCOLS: {
759 		uint8_t subtype;
760 
761 		if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype)) {
762 			m_freem(m);
763 			return;
764 		}
765 
766 		m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype);
767 		switch (subtype) {
768 #if NAGR > 0
769 		case SLOWPROTOCOLS_SUBTYPE_LACP:
770 			if (ifp->if_agrprivate) {
771 				ieee8023ad_lacp_input(ifp, m);
772 				return;
773 			}
774 			break;
775 
776 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
777 			if (ifp->if_agrprivate) {
778 				ieee8023ad_marker_input(ifp, m);
779 				return;
780 			}
781 			break;
782 #endif
783 
784 		default:
785 			if (subtype == 0 || subtype > 10) {
786 				/* illegal value */
787 				m_freem(m);
788 				return;
789 			}
790 			/* unknown subtype */
791 			break;
792 		}
793 	}
794 	/* FALLTHROUGH */
795 	default:
796 		if (m->m_flags & M_PROMISC) {
797 			m_freem(m);
798 			return;
799 		}
800 	}
801 
802 	/* If the CRC is still on the packet, trim it off. */
803 	if (m->m_flags & M_HASFCS) {
804 		m_adj(m, -ETHER_CRC_LEN);
805 		m->m_flags &= ~M_HASFCS;
806 	}
807 
808 	if (etype > ETHERMTU + sizeof(struct ether_header)) {
809 		/* Strip off the Ethernet header. */
810 		m_adj(m, ehlen);
811 
812 		switch (etype) {
813 #ifdef INET
814 		case ETHERTYPE_IP:
815 #ifdef GATEWAY
816 			if (ipflow_fastforward(m))
817 				return;
818 #endif
819 			pktq = ip_pktq;
820 			break;
821 
822 		case ETHERTYPE_ARP:
823 			isr = NETISR_ARP;
824 			inq = &arpintrq;
825 			break;
826 
827 		case ETHERTYPE_REVARP:
828 			revarpinput(m);	/* XXX queue? */
829 			return;
830 #endif
831 
832 #ifdef INET6
833 		case ETHERTYPE_IPV6:
834 			if (__predict_false(!in6_present)) {
835 				m_freem(m);
836 				return;
837 			}
838 #ifdef GATEWAY
839 			if (ip6flow_fastforward(&m))
840 				return;
841 #endif
842 			pktq = ip6_pktq;
843 			break;
844 #endif
845 
846 #ifdef NETATALK
847 		case ETHERTYPE_ATALK:
848 			isr = NETISR_ATALK;
849 			inq = &atintrq1;
850 			break;
851 
852 		case ETHERTYPE_AARP:
853 			aarpinput(ifp, m); /* XXX queue? */
854 			return;
855 #endif
856 
857 #ifdef MPLS
858 		case ETHERTYPE_MPLS:
859 			isr = NETISR_MPLS;
860 			inq = &mplsintrq;
861 			break;
862 #endif
863 
864 		default:
865 			m_freem(m);
866 			return;
867 		}
868 	} else {
869 		KASSERT(ehlen == sizeof(*eh));
870 #if defined (LLC) || defined (NETATALK)
871 		if (m->m_len < sizeof(*eh) + sizeof(struct llc)) {
872 			goto dropanyway;
873 		}
874 		l = (struct llc *)(eh+1);
875 
876 		switch (l->llc_dsap) {
877 #ifdef NETATALK
878 		case LLC_SNAP_LSAP:
879 			switch (l->llc_control) {
880 			case LLC_UI:
881 				if (l->llc_ssap != LLC_SNAP_LSAP) {
882 					goto dropanyway;
883 				}
884 
885 				if (memcmp(&(l->llc_snap_org_code)[0],
886 				    at_org_code, sizeof(at_org_code)) == 0 &&
887 				    ntohs(l->llc_snap_ether_type) ==
888 				    ETHERTYPE_ATALK) {
889 					inq = &atintrq2;
890 					m_adj(m, sizeof(struct ether_header)
891 					    + sizeof(struct llc));
892 					isr = NETISR_ATALK;
893 					break;
894 				}
895 
896 				if (memcmp(&(l->llc_snap_org_code)[0],
897 				    aarp_org_code,
898 				    sizeof(aarp_org_code)) == 0 &&
899 				    ntohs(l->llc_snap_ether_type) ==
900 				    ETHERTYPE_AARP) {
901 					m_adj(m, sizeof(struct ether_header)
902 					    + sizeof(struct llc));
903 					aarpinput(ifp, m); /* XXX queue? */
904 					return;
905 				}
906 
907 			default:
908 				goto dropanyway;
909 			}
910 			break;
911 #endif
912 		dropanyway:
913 		default:
914 			m_freem(m);
915 			return;
916 		}
917 #else /* LLC || NETATALK */
918 		m_freem(m);
919 		return;
920 #endif /* LLC || NETATALK */
921 	}
922 
923 	if (__predict_true(pktq)) {
924 #ifdef NET_MPSAFE
925 		const u_int h = curcpu()->ci_index;
926 #else
927 		const uint32_t h = pktq_rps_hash(m);
928 #endif
929 		if (__predict_false(!pktq_enqueue(pktq, m, h))) {
930 			m_freem(m);
931 		}
932 		return;
933 	}
934 
935 	if (__predict_false(!inq)) {
936 		/* Should not happen. */
937 		m_freem(m);
938 		return;
939 	}
940 
941 	IFQ_LOCK(inq);
942 	if (IF_QFULL(inq)) {
943 		IF_DROP(inq);
944 		IFQ_UNLOCK(inq);
945 		m_freem(m);
946 	} else {
947 		IF_ENQUEUE(inq, m);
948 		IFQ_UNLOCK(inq);
949 		schednetisr(isr);
950 	}
951 }
952 
953 /*
954  * Convert Ethernet address to printable (loggable) representation.
955  */
956 char *
957 ether_sprintf(const u_char *ap)
958 {
959 	static char etherbuf[3 * ETHER_ADDR_LEN];
960 	return ether_snprintf(etherbuf, sizeof(etherbuf), ap);
961 }
962 
963 char *
964 ether_snprintf(char *buf, size_t len, const u_char *ap)
965 {
966 	char *cp = buf;
967 	size_t i;
968 
969 	for (i = 0; i < len / 3; i++) {
970 		*cp++ = hexdigits[*ap >> 4];
971 		*cp++ = hexdigits[*ap++ & 0xf];
972 		*cp++ = ':';
973 	}
974 	*--cp = '\0';
975 	return buf;
976 }
977 
978 /*
979  * Perform common duties while attaching to interface list
980  */
981 void
982 ether_ifattach(struct ifnet *ifp, const uint8_t *lla)
983 {
984 	struct ethercom *ec = (struct ethercom *)ifp;
985 
986 	ifp->if_type = IFT_ETHER;
987 	ifp->if_hdrlen = ETHER_HDR_LEN;
988 	ifp->if_dlt = DLT_EN10MB;
989 	ifp->if_mtu = ETHERMTU;
990 	ifp->if_output = ether_output;
991 	ifp->_if_input = ether_input;
992 	if (ifp->if_baudrate == 0)
993 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
994 
995 	if (lla != NULL)
996 		if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla));
997 
998 	LIST_INIT(&ec->ec_multiaddrs);
999 	SIMPLEQ_INIT(&ec->ec_vids);
1000 	ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
1001 	ec->ec_flags = 0;
1002 	ifp->if_broadcastaddr = etherbroadcastaddr;
1003 	bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header));
1004 #ifdef MBUFTRACE
1005 	strlcpy(ec->ec_tx_mowner.mo_name, ifp->if_xname,
1006 	    sizeof(ec->ec_tx_mowner.mo_name));
1007 	strlcpy(ec->ec_tx_mowner.mo_descr, "tx",
1008 	    sizeof(ec->ec_tx_mowner.mo_descr));
1009 	strlcpy(ec->ec_rx_mowner.mo_name, ifp->if_xname,
1010 	    sizeof(ec->ec_rx_mowner.mo_name));
1011 	strlcpy(ec->ec_rx_mowner.mo_descr, "rx",
1012 	    sizeof(ec->ec_rx_mowner.mo_descr));
1013 	MOWNER_ATTACH(&ec->ec_tx_mowner);
1014 	MOWNER_ATTACH(&ec->ec_rx_mowner);
1015 	ifp->if_mowner = &ec->ec_tx_mowner;
1016 #endif
1017 }
1018 
1019 void
1020 ether_ifdetach(struct ifnet *ifp)
1021 {
1022 	struct ethercom *ec = (void *) ifp;
1023 	struct ether_multi *enm;
1024 
1025 	IFNET_ASSERT_UNLOCKED(ifp);
1026 	/*
1027 	 * Prevent further calls to ioctl (for example turning off
1028 	 * promiscuous mode from the bridge code), which eventually can
1029 	 * call if_init() which can cause panics because the interface
1030 	 * is in the process of being detached. Return device not configured
1031 	 * instead.
1032 	 */
1033 	ifp->if_ioctl = (int (*)(struct ifnet *, u_long, void *))enxio;
1034 
1035 #if NBRIDGE > 0
1036 	if (ifp->if_bridge)
1037 		bridge_ifdetach(ifp);
1038 #endif
1039 	bpf_detach(ifp);
1040 #if NVLAN > 0
1041 	if (ec->ec_nvlans)
1042 		vlan_ifdetach(ifp);
1043 #endif
1044 
1045 	ETHER_LOCK(ec);
1046 	KASSERT(ec->ec_nvlans == 0);
1047 	while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) {
1048 		LIST_REMOVE(enm, enm_list);
1049 		kmem_free(enm, sizeof(*enm));
1050 		ec->ec_multicnt--;
1051 	}
1052 	ETHER_UNLOCK(ec);
1053 
1054 	mutex_obj_free(ec->ec_lock);
1055 	ec->ec_lock = NULL;
1056 
1057 	ifp->if_mowner = NULL;
1058 	MOWNER_DETACH(&ec->ec_rx_mowner);
1059 	MOWNER_DETACH(&ec->ec_tx_mowner);
1060 }
1061 
1062 #if 0
1063 /*
1064  * This is for reference.  We have a table-driven version
1065  * of the little-endian crc32 generator, which is faster
1066  * than the double-loop.
1067  */
1068 uint32_t
1069 ether_crc32_le(const uint8_t *buf, size_t len)
1070 {
1071 	uint32_t c, crc, carry;
1072 	size_t i, j;
1073 
1074 	crc = 0xffffffffU;	/* initial value */
1075 
1076 	for (i = 0; i < len; i++) {
1077 		c = buf[i];
1078 		for (j = 0; j < 8; j++) {
1079 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1080 			crc >>= 1;
1081 			c >>= 1;
1082 			if (carry)
1083 				crc = (crc ^ ETHER_CRC_POLY_LE);
1084 		}
1085 	}
1086 
1087 	return (crc);
1088 }
1089 #else
1090 uint32_t
1091 ether_crc32_le(const uint8_t *buf, size_t len)
1092 {
1093 	static const uint32_t crctab[] = {
1094 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1095 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1096 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1097 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1098 	};
1099 	uint32_t crc;
1100 	size_t i;
1101 
1102 	crc = 0xffffffffU;	/* initial value */
1103 
1104 	for (i = 0; i < len; i++) {
1105 		crc ^= buf[i];
1106 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1107 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1108 	}
1109 
1110 	return (crc);
1111 }
1112 #endif
1113 
1114 uint32_t
1115 ether_crc32_be(const uint8_t *buf, size_t len)
1116 {
1117 	uint32_t c, crc, carry;
1118 	size_t i, j;
1119 
1120 	crc = 0xffffffffU;	/* initial value */
1121 
1122 	for (i = 0; i < len; i++) {
1123 		c = buf[i];
1124 		for (j = 0; j < 8; j++) {
1125 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1126 			crc <<= 1;
1127 			c >>= 1;
1128 			if (carry)
1129 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1130 		}
1131 	}
1132 
1133 	return (crc);
1134 }
1135 
1136 #ifdef INET
1137 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] =
1138     { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 };
1139 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] =
1140     { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff };
1141 #endif
1142 #ifdef INET6
1143 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] =
1144     { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 };
1145 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] =
1146     { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff };
1147 #endif
1148 
1149 /*
1150  * ether_aton implementation, not using a static buffer.
1151  */
1152 int
1153 ether_aton_r(u_char *dest, size_t len, const char *str)
1154 {
1155 	const u_char *cp = (const void *)str;
1156 	u_char *ep;
1157 
1158 #define atox(c)	(((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10))
1159 
1160 	if (len < ETHER_ADDR_LEN)
1161 		return ENOSPC;
1162 
1163 	ep = dest + ETHER_ADDR_LEN;
1164 
1165 	while (*cp) {
1166 		if (!isxdigit(*cp))
1167 			return EINVAL;
1168 
1169 		*dest = atox(*cp);
1170 		cp++;
1171 		if (isxdigit(*cp)) {
1172 			*dest = (*dest << 4) | atox(*cp);
1173 			cp++;
1174 		}
1175 		dest++;
1176 
1177 		if (dest == ep)
1178 			return (*cp == '\0') ? 0 : ENAMETOOLONG;
1179 
1180 		switch (*cp) {
1181 		case ':':
1182 		case '-':
1183 		case '.':
1184 			cp++;
1185 			break;
1186 		}
1187 	}
1188 	return ENOBUFS;
1189 }
1190 
1191 /*
1192  * Convert a sockaddr into an Ethernet address or range of Ethernet
1193  * addresses.
1194  */
1195 int
1196 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN],
1197     uint8_t addrhi[ETHER_ADDR_LEN])
1198 {
1199 #ifdef INET
1200 	const struct sockaddr_in *sin;
1201 #endif
1202 #ifdef INET6
1203 	const struct sockaddr_in6 *sin6;
1204 #endif
1205 
1206 	switch (sa->sa_family) {
1207 
1208 	case AF_UNSPEC:
1209 		memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN);
1210 		memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1211 		break;
1212 
1213 #ifdef INET
1214 	case AF_INET:
1215 		sin = satocsin(sa);
1216 		if (sin->sin_addr.s_addr == INADDR_ANY) {
1217 			/*
1218 			 * An IP address of INADDR_ANY means listen to
1219 			 * or stop listening to all of the Ethernet
1220 			 * multicast addresses used for IP.
1221 			 * (This is for the sake of IP multicast routers.)
1222 			 */
1223 			memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN);
1224 			memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN);
1225 		} else {
1226 			ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo);
1227 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1228 		}
1229 		break;
1230 #endif
1231 #ifdef INET6
1232 	case AF_INET6:
1233 		sin6 = satocsin6(sa);
1234 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1235 			/*
1236 			 * An IP6 address of 0 means listen to or stop
1237 			 * listening to all of the Ethernet multicast
1238 			 * address used for IP6.
1239 			 * (This is used for multicast routers.)
1240 			 */
1241 			memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN);
1242 			memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN);
1243 		} else {
1244 			ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo);
1245 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1246 		}
1247 		break;
1248 #endif
1249 
1250 	default:
1251 		return EAFNOSUPPORT;
1252 	}
1253 	return 0;
1254 }
1255 
1256 /*
1257  * Add an Ethernet multicast address or range of addresses to the list for a
1258  * given interface.
1259  */
1260 int
1261 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec)
1262 {
1263 	struct ether_multi *enm, *_enm;
1264 	u_char addrlo[ETHER_ADDR_LEN];
1265 	u_char addrhi[ETHER_ADDR_LEN];
1266 	int error = 0;
1267 
1268 	/* Allocate out of lock */
1269 	enm = kmem_alloc(sizeof(*enm), KM_SLEEP);
1270 
1271 	ETHER_LOCK(ec);
1272 	error = ether_multiaddr(sa, addrlo, addrhi);
1273 	if (error != 0)
1274 		goto out;
1275 
1276 	/*
1277 	 * Verify that we have valid Ethernet multicast addresses.
1278 	 */
1279 	if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) {
1280 		error = EINVAL;
1281 		goto out;
1282 	}
1283 
1284 	/*
1285 	 * See if the address range is already in the list.
1286 	 */
1287 	_enm = ether_lookup_multi(addrlo, addrhi, ec);
1288 	if (_enm != NULL) {
1289 		/*
1290 		 * Found it; just increment the reference count.
1291 		 */
1292 		++_enm->enm_refcount;
1293 		error = 0;
1294 		goto out;
1295 	}
1296 
1297 	/*
1298 	 * Link a new multicast record into the interface's multicast list.
1299 	 */
1300 	memcpy(enm->enm_addrlo, addrlo, ETHER_ADDR_LEN);
1301 	memcpy(enm->enm_addrhi, addrhi, ETHER_ADDR_LEN);
1302 	enm->enm_refcount = 1;
1303 	LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list);
1304 	ec->ec_multicnt++;
1305 
1306 	/*
1307 	 * Return ENETRESET to inform the driver that the list has changed
1308 	 * and its reception filter should be adjusted accordingly.
1309 	 */
1310 	error = ENETRESET;
1311 	enm = NULL;
1312 
1313 out:
1314 	ETHER_UNLOCK(ec);
1315 	if (enm != NULL)
1316 		kmem_free(enm, sizeof(*enm));
1317 	return error;
1318 }
1319 
1320 /*
1321  * Delete a multicast address record.
1322  */
1323 int
1324 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec)
1325 {
1326 	struct ether_multi *enm;
1327 	u_char addrlo[ETHER_ADDR_LEN];
1328 	u_char addrhi[ETHER_ADDR_LEN];
1329 	int error;
1330 
1331 	ETHER_LOCK(ec);
1332 	error = ether_multiaddr(sa, addrlo, addrhi);
1333 	if (error != 0)
1334 		goto error;
1335 
1336 	/*
1337 	 * Look up the address in our list.
1338 	 */
1339 	enm = ether_lookup_multi(addrlo, addrhi, ec);
1340 	if (enm == NULL) {
1341 		error = ENXIO;
1342 		goto error;
1343 	}
1344 	if (--enm->enm_refcount != 0) {
1345 		/*
1346 		 * Still some claims to this record.
1347 		 */
1348 		error = 0;
1349 		goto error;
1350 	}
1351 
1352 	/*
1353 	 * No remaining claims to this record; unlink and free it.
1354 	 */
1355 	LIST_REMOVE(enm, enm_list);
1356 	ec->ec_multicnt--;
1357 	ETHER_UNLOCK(ec);
1358 	kmem_free(enm, sizeof(*enm));
1359 
1360 	/*
1361 	 * Return ENETRESET to inform the driver that the list has changed
1362 	 * and its reception filter should be adjusted accordingly.
1363 	 */
1364 	return ENETRESET;
1365 
1366 error:
1367 	ETHER_UNLOCK(ec);
1368 	return error;
1369 }
1370 
1371 void
1372 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb)
1373 {
1374 	ec->ec_ifflags_cb = cb;
1375 }
1376 
1377 void
1378 ether_set_vlan_cb(struct ethercom *ec, ether_vlancb_t cb)
1379 {
1380 
1381 	ec->ec_vlan_cb = cb;
1382 }
1383 
1384 static int
1385 ether_ioctl_reinit(struct ethercom *ec)
1386 {
1387 	struct ifnet *ifp = &ec->ec_if;
1388 	int error;
1389 
1390 	switch (ifp->if_flags & (IFF_UP | IFF_RUNNING)) {
1391 	case IFF_RUNNING:
1392 		/*
1393 		 * If interface is marked down and it is running,
1394 		 * then stop and disable it.
1395 		 */
1396 		(*ifp->if_stop)(ifp, 1);
1397 		break;
1398 	case IFF_UP:
1399 		/*
1400 		 * If interface is marked up and it is stopped, then
1401 		 * start it.
1402 		 */
1403 		return (*ifp->if_init)(ifp);
1404 	case IFF_UP | IFF_RUNNING:
1405 		error = 0;
1406 		if (ec->ec_ifflags_cb != NULL) {
1407 			error = (*ec->ec_ifflags_cb)(ec);
1408 			if (error == ENETRESET) {
1409 				/*
1410 				 * Reset the interface to pick up
1411 				 * changes in any other flags that
1412 				 * affect the hardware state.
1413 				 */
1414 				return (*ifp->if_init)(ifp);
1415 			}
1416 		} else
1417 			error = (*ifp->if_init)(ifp);
1418 		return error;
1419 	case 0:
1420 		break;
1421 	}
1422 
1423 	return 0;
1424 }
1425 
1426 /*
1427  * Common ioctls for Ethernet interfaces.  Note, we must be
1428  * called at splnet().
1429  */
1430 int
1431 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data)
1432 {
1433 	struct ethercom *ec = (void *)ifp;
1434 	struct eccapreq *eccr;
1435 	struct ifreq *ifr = (struct ifreq *)data;
1436 	struct if_laddrreq *iflr = data;
1437 	const struct sockaddr_dl *sdl;
1438 	static const uint8_t zero[ETHER_ADDR_LEN];
1439 	int error;
1440 
1441 	switch (cmd) {
1442 	case SIOCINITIFADDR:
1443 	    {
1444 		struct ifaddr *ifa = (struct ifaddr *)data;
1445 		if (ifa->ifa_addr->sa_family != AF_LINK
1446 		    && (ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
1447 		       (IFF_UP | IFF_RUNNING)) {
1448 			ifp->if_flags |= IFF_UP;
1449 			if ((error = (*ifp->if_init)(ifp)) != 0)
1450 				return error;
1451 		}
1452 #ifdef INET
1453 		if (ifa->ifa_addr->sa_family == AF_INET)
1454 			arp_ifinit(ifp, ifa);
1455 #endif
1456 		return 0;
1457 	    }
1458 
1459 	case SIOCSIFMTU:
1460 	    {
1461 		int maxmtu;
1462 
1463 		if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU)
1464 			maxmtu = ETHERMTU_JUMBO;
1465 		else
1466 			maxmtu = ETHERMTU;
1467 
1468 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu)
1469 			return EINVAL;
1470 		else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET)
1471 			return error;
1472 		else if (ifp->if_flags & IFF_UP) {
1473 			/* Make sure the device notices the MTU change. */
1474 			return (*ifp->if_init)(ifp);
1475 		} else
1476 			return 0;
1477 	    }
1478 
1479 	case SIOCSIFFLAGS:
1480 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
1481 			return error;
1482 		return ether_ioctl_reinit(ec);
1483 	case SIOCGIFFLAGS:
1484 		error = ifioctl_common(ifp, cmd, data);
1485 		if (error == 0) {
1486 			/* Set IFF_ALLMULTI for backcompat */
1487 			ifr->ifr_flags |= (ec->ec_flags & ETHER_F_ALLMULTI) ?
1488 			    IFF_ALLMULTI : 0;
1489 		}
1490 		return error;
1491 	case SIOCGETHERCAP:
1492 		eccr = (struct eccapreq *)data;
1493 		eccr->eccr_capabilities = ec->ec_capabilities;
1494 		eccr->eccr_capenable = ec->ec_capenable;
1495 		return 0;
1496 	case SIOCSETHERCAP:
1497 		eccr = (struct eccapreq *)data;
1498 		if ((eccr->eccr_capenable & ~ec->ec_capabilities) != 0)
1499 			return EINVAL;
1500 		if (eccr->eccr_capenable == ec->ec_capenable)
1501 			return 0;
1502 #if 0 /* notyet */
1503 		ec->ec_capenable = (ec->ec_capenable & ETHERCAP_CANTCHANGE)
1504 		    | (eccr->eccr_capenable & ~ETHERCAP_CANTCHANGE);
1505 #else
1506 		ec->ec_capenable = eccr->eccr_capenable;
1507 #endif
1508 		return ether_ioctl_reinit(ec);
1509 	case SIOCADDMULTI:
1510 		return ether_addmulti(ifreq_getaddr(cmd, ifr), ec);
1511 	case SIOCDELMULTI:
1512 		return ether_delmulti(ifreq_getaddr(cmd, ifr), ec);
1513 	case SIOCSIFMEDIA:
1514 	case SIOCGIFMEDIA:
1515 		if (ec->ec_mii != NULL)
1516 			return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media,
1517 			    cmd);
1518 		else if (ec->ec_ifmedia != NULL)
1519 			return ifmedia_ioctl(ifp, ifr, ec->ec_ifmedia, cmd);
1520 		else
1521 			return ENOTTY;
1522 		break;
1523 	case SIOCALIFADDR:
1524 		sdl = satocsdl(sstocsa(&iflr->addr));
1525 		if (sdl->sdl_family != AF_LINK)
1526 			;
1527 		else if (ETHER_IS_MULTICAST(CLLADDR(sdl)))
1528 			return EINVAL;
1529 		else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0)
1530 			return EINVAL;
1531 		/*FALLTHROUGH*/
1532 	default:
1533 		return ifioctl_common(ifp, cmd, data);
1534 	}
1535 	return 0;
1536 }
1537 
1538 /*
1539  * Enable/disable passing VLAN packets if the parent interface supports it.
1540  * Return:
1541  * 	 0: Ok
1542  *	-1: Parent interface does not support vlans
1543  *	>0: Error
1544  */
1545 int
1546 ether_enable_vlan_mtu(struct ifnet *ifp)
1547 {
1548 	int error;
1549 	struct ethercom *ec = (void *)ifp;
1550 
1551 	/* Parent does not support VLAN's */
1552 	if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0)
1553 		return -1;
1554 
1555 	/*
1556 	 * Parent supports the VLAN_MTU capability,
1557 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames;
1558 	 * enable it.
1559 	 */
1560 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1561 
1562 	/* Interface is down, defer for later */
1563 	if ((ifp->if_flags & IFF_UP) == 0)
1564 		return 0;
1565 
1566 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1567 		return 0;
1568 
1569 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1570 	return error;
1571 }
1572 
1573 int
1574 ether_disable_vlan_mtu(struct ifnet *ifp)
1575 {
1576 	int error;
1577 	struct ethercom *ec = (void *)ifp;
1578 
1579 	/* We still have VLAN's, defer for later */
1580 	if (ec->ec_nvlans != 0)
1581 		return 0;
1582 
1583 	/* Parent does not support VLAB's, nothing to do. */
1584 	if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0)
1585 		return -1;
1586 
1587 	/*
1588 	 * Disable Tx/Rx of VLAN-sized frames.
1589 	 */
1590 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1591 
1592 	/* Interface is down, defer for later */
1593 	if ((ifp->if_flags & IFF_UP) == 0)
1594 		return 0;
1595 
1596 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1597 		return 0;
1598 
1599 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1600 	return error;
1601 }
1602 
1603 static int
1604 ether_multicast_sysctl(SYSCTLFN_ARGS)
1605 {
1606 	struct ether_multi *enm;
1607 	struct ifnet *ifp;
1608 	struct ethercom *ec;
1609 	int error = 0;
1610 	size_t written;
1611 	struct psref psref;
1612 	int bound;
1613 	unsigned int multicnt;
1614 	struct ether_multi_sysctl *addrs;
1615 	int i;
1616 
1617 	if (namelen != 1)
1618 		return EINVAL;
1619 
1620 	bound = curlwp_bind();
1621 	ifp = if_get_byindex(name[0], &psref);
1622 	if (ifp == NULL) {
1623 		error = ENODEV;
1624 		goto out;
1625 	}
1626 	if (ifp->if_type != IFT_ETHER) {
1627 		if_put(ifp, &psref);
1628 		*oldlenp = 0;
1629 		goto out;
1630 	}
1631 	ec = (struct ethercom *)ifp;
1632 
1633 	if (oldp == NULL) {
1634 		if_put(ifp, &psref);
1635 		*oldlenp = ec->ec_multicnt * sizeof(*addrs);
1636 		goto out;
1637 	}
1638 
1639 	/*
1640 	 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which
1641 	 * is sleepable, while holding it. Copy data to a local buffer first
1642 	 * with the lock taken and then call sysctl_copyout without holding it.
1643 	 */
1644 retry:
1645 	multicnt = ec->ec_multicnt;
1646 
1647 	if (multicnt == 0) {
1648 		if_put(ifp, &psref);
1649 		*oldlenp = 0;
1650 		goto out;
1651 	}
1652 
1653 	addrs = kmem_zalloc(sizeof(*addrs) * multicnt, KM_SLEEP);
1654 
1655 	ETHER_LOCK(ec);
1656 	if (multicnt != ec->ec_multicnt) {
1657 		/* The number of multicast addresses has changed */
1658 		ETHER_UNLOCK(ec);
1659 		kmem_free(addrs, sizeof(*addrs) * multicnt);
1660 		goto retry;
1661 	}
1662 
1663 	i = 0;
1664 	LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) {
1665 		struct ether_multi_sysctl *addr = &addrs[i];
1666 		addr->enm_refcount = enm->enm_refcount;
1667 		memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN);
1668 		memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN);
1669 		i++;
1670 	}
1671 	ETHER_UNLOCK(ec);
1672 
1673 	error = 0;
1674 	written = 0;
1675 	for (i = 0; i < multicnt; i++) {
1676 		struct ether_multi_sysctl *addr = &addrs[i];
1677 
1678 		if (written + sizeof(*addr) > *oldlenp)
1679 			break;
1680 		error = sysctl_copyout(l, addr, oldp, sizeof(*addr));
1681 		if (error)
1682 			break;
1683 		written += sizeof(*addr);
1684 		oldp = (char *)oldp + sizeof(*addr);
1685 	}
1686 	kmem_free(addrs, sizeof(*addrs) * multicnt);
1687 
1688 	if_put(ifp, &psref);
1689 
1690 	*oldlenp = written;
1691 out:
1692 	curlwp_bindx(bound);
1693 	return error;
1694 }
1695 
1696 static void
1697 ether_sysctl_setup(struct sysctllog **clog)
1698 {
1699 	const struct sysctlnode *rnode = NULL;
1700 
1701 	sysctl_createv(clog, 0, NULL, &rnode,
1702 		       CTLFLAG_PERMANENT,
1703 		       CTLTYPE_NODE, "ether",
1704 		       SYSCTL_DESCR("Ethernet-specific information"),
1705 		       NULL, 0, NULL, 0,
1706 		       CTL_NET, CTL_CREATE, CTL_EOL);
1707 
1708 	sysctl_createv(clog, 0, &rnode, NULL,
1709 		       CTLFLAG_PERMANENT,
1710 		       CTLTYPE_NODE, "multicast",
1711 		       SYSCTL_DESCR("multicast addresses"),
1712 		       ether_multicast_sysctl, 0, NULL, 0,
1713 		       CTL_CREATE, CTL_EOL);
1714 }
1715 
1716 void
1717 etherinit(void)
1718 {
1719 
1720 	mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET);
1721 	ether_sysctl_setup(NULL);
1722 }
1723