xref: /netbsd-src/sys/net/if_ethersubr.c (revision 7e30e94394d0994ab9534f68a8f91665045c91ce)
1 /*	$NetBSD: if_ethersubr.c,v 1.239 2017/02/21 03:59:31 ozaki-r Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (c) 1982, 1989, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)if_ethersubr.c	8.2 (Berkeley) 4/4/96
61  */
62 
63 #include <sys/cdefs.h>
64 __KERNEL_RCSID(0, "$NetBSD: if_ethersubr.c,v 1.239 2017/02/21 03:59:31 ozaki-r Exp $");
65 
66 #ifdef _KERNEL_OPT
67 #include "opt_inet.h"
68 #include "opt_atalk.h"
69 #include "opt_mbuftrace.h"
70 #include "opt_mpls.h"
71 #include "opt_gateway.h"
72 #include "opt_pppoe.h"
73 #include "opt_net_mpsafe.h"
74 #endif
75 
76 #include "vlan.h"
77 #include "pppoe.h"
78 #include "bridge.h"
79 #include "arp.h"
80 #include "agr.h"
81 
82 #include <sys/sysctl.h>
83 #include <sys/mbuf.h>
84 #include <sys/mutex.h>
85 #include <sys/ioctl.h>
86 #include <sys/errno.h>
87 #include <sys/device.h>
88 #include <sys/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 
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 static struct timeval bigpktppslim_last;
169 static int bigpktppslim = 2;	/* XXX */
170 static int bigpktpps_count;
171 static kmutex_t bigpktpps_lock __cacheline_aligned;
172 
173 
174 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] =
175     { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
176 const uint8_t ethermulticastaddr_slowprotocols[ETHER_ADDR_LEN] =
177     { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x02 };
178 #define senderr(e) { error = (e); goto bad;}
179 
180 static	int ether_output(struct ifnet *, struct mbuf *,
181 	    const struct sockaddr *, const struct rtentry *);
182 
183 /*
184  * Ethernet output routine.
185  * Encapsulate a packet of type family for the local net.
186  * Assumes that ifp is actually pointer to ethercom structure.
187  */
188 static int
189 ether_output(struct ifnet * const ifp0, struct mbuf * const m0,
190 	const struct sockaddr * const dst,
191 	const struct rtentry *rt)
192 {
193 	uint16_t etype = 0;
194 	int error = 0, hdrcmplt = 0;
195  	uint8_t esrc[6], edst[6];
196 	struct mbuf *m = m0;
197 	struct mbuf *mcopy = NULL;
198 	struct ether_header *eh;
199 	struct ifnet *ifp = ifp0;
200 #ifdef INET
201 	struct arphdr *ah;
202 #endif /* INET */
203 #ifdef NETATALK
204 	struct at_ifaddr *aa;
205 #endif /* NETATALK */
206 
207 	/*
208 	 * some paths such as carp_output() call ethr_output() with "ifp"
209 	 * argument as other than ether ifnet.
210 	 */
211 	KASSERT(ifp->if_output != ether_output
212 	    || ifp->if_extflags & IFEF_OUTPUT_MPSAFE);
213 
214 #ifdef MBUFTRACE
215 	m_claimm(m, ifp->if_mowner);
216 #endif
217 
218 #if NCARP > 0
219 	if (ifp->if_type == IFT_CARP) {
220 		struct ifaddr *ifa;
221 		int s = pserialize_read_enter();
222 
223 		/* loop back if this is going to the carp interface */
224 		if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP &&
225 		    (ifa = ifa_ifwithaddr(dst)) != NULL) {
226 			if (ifa->ifa_ifp == ifp0) {
227 				pserialize_read_exit(s);
228 				return looutput(ifp0, m, dst, rt);
229 			}
230 		}
231 		pserialize_read_exit(s);
232 
233 		ifp = ifp->if_carpdev;
234 		/* ac = (struct arpcom *)ifp; */
235 
236 		if ((ifp0->if_flags & (IFF_UP|IFF_RUNNING)) !=
237 		    (IFF_UP|IFF_RUNNING))
238 			senderr(ENETDOWN);
239 	}
240 #endif /* NCARP > 0 */
241 
242 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
243 		senderr(ENETDOWN);
244 
245 	switch (dst->sa_family) {
246 
247 #ifdef INET
248 	case AF_INET:
249 #ifndef NET_MPSAFE
250 		KERNEL_LOCK(1, NULL);
251 #endif
252 		if (m->m_flags & M_BCAST)
253 			(void)memcpy(edst, etherbroadcastaddr, sizeof(edst));
254 		else if (m->m_flags & M_MCAST)
255 			ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst);
256 		else if ((error = arpresolve(ifp, rt, m, dst, edst,
257 		    sizeof(edst))) != 0) {
258 #ifndef NET_MPSAFE
259 			KERNEL_UNLOCK_ONE(NULL);
260 #endif
261 			return error == EWOULDBLOCK ? 0 : error;
262 		}
263 #ifndef NET_MPSAFE
264 		KERNEL_UNLOCK_ONE(NULL);
265 #endif
266 		/* If broadcasting on a simplex interface, loopback a copy */
267 		if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
268 			mcopy = m_copy(m, 0, (int)M_COPYALL);
269 		etype = htons(ETHERTYPE_IP);
270 		break;
271 
272 	case AF_ARP:
273 		ah = mtod(m, struct arphdr *);
274 		if (m->m_flags & M_BCAST)
275 			(void)memcpy(edst, etherbroadcastaddr, sizeof(edst));
276 		else {
277 			void *tha = ar_tha(ah);
278 
279 			if (tha == NULL) {
280 				/* fake with ARPHDR_IEEE1394 */
281 				m_freem(m);
282 				return 0;
283 			}
284 			memcpy(edst, tha, sizeof(edst));
285 		}
286 
287 		ah->ar_hrd = htons(ARPHRD_ETHER);
288 
289 		switch (ntohs(ah->ar_op)) {
290 		case ARPOP_REVREQUEST:
291 		case ARPOP_REVREPLY:
292 			etype = htons(ETHERTYPE_REVARP);
293 			break;
294 
295 		case ARPOP_REQUEST:
296 		case ARPOP_REPLY:
297 		default:
298 			etype = htons(ETHERTYPE_ARP);
299 		}
300 
301 		break;
302 #endif
303 #ifdef INET6
304 	case AF_INET6:
305 		if (m->m_flags & M_BCAST)
306 			(void)memcpy(edst, etherbroadcastaddr, sizeof(edst));
307 		else if (m->m_flags & M_MCAST) {
308 			ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr,
309 			    edst);
310 		} else {
311 			error = nd6_resolve(ifp, rt, m, dst, edst,
312 			    sizeof(edst));
313 			if (error != 0)
314 				return error == EWOULDBLOCK ? 0 : error;
315 		}
316 		etype = htons(ETHERTYPE_IPV6);
317 		break;
318 #endif
319 #ifdef NETATALK
320     case AF_APPLETALK: {
321 		struct ifaddr *ifa;
322 		int s;
323 
324 		KERNEL_LOCK(1, NULL);
325 		if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) {
326 #ifdef NETATALKDEBUG
327 			printf("aarpresolv failed\n");
328 #endif /* NETATALKDEBUG */
329 			KERNEL_UNLOCK_ONE(NULL);
330 			return (0);
331 		}
332 		/*
333 		 * ifaddr is the first thing in at_ifaddr
334 		 */
335 		s = pserialize_read_enter();
336 		ifa = at_ifawithnet((const struct sockaddr_at *)dst, ifp);
337 		if (ifa == NULL) {
338 			pserialize_read_exit(s);
339 			KERNEL_UNLOCK_ONE(NULL);
340 			goto bad;
341 		}
342 		aa = (struct at_ifaddr *)ifa;
343 
344 		/*
345 		 * In the phase 2 case, we need to prepend an mbuf for the
346 		 * llc header.  Since we must preserve the value of m,
347 		 * which is passed to us by value, we m_copy() the first
348 		 * mbuf, and use it for our llc header.
349 		 */
350 		if (aa->aa_flags & AFA_PHASE2) {
351 			struct llc llc;
352 
353 			M_PREPEND(m, sizeof(struct llc), M_DONTWAIT);
354 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
355 			llc.llc_control = LLC_UI;
356 			memcpy(llc.llc_snap_org_code, at_org_code,
357 			    sizeof(llc.llc_snap_org_code));
358 			llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK);
359 			memcpy(mtod(m, void *), &llc, sizeof(struct llc));
360 		} else {
361 			etype = htons(ETHERTYPE_ATALK);
362 		}
363 		pserialize_read_exit(s);
364 		KERNEL_UNLOCK_ONE(NULL);
365 		break;
366 	    }
367 #endif /* NETATALK */
368 	case pseudo_AF_HDRCMPLT:
369 		hdrcmplt = 1;
370 		memcpy(esrc,
371 		    ((const struct ether_header *)dst->sa_data)->ether_shost,
372 		    sizeof(esrc));
373 		/* FALLTHROUGH */
374 
375 	case AF_UNSPEC:
376  		memcpy(edst,
377 		    ((const struct ether_header *)dst->sa_data)->ether_dhost,
378 		    sizeof(edst));
379 		/* AF_UNSPEC doesn't swap the byte order of the ether_type. */
380 		etype = ((const struct ether_header *)dst->sa_data)->ether_type;
381 		break;
382 
383 	default:
384 		printf("%s: can't handle af%d\n", ifp->if_xname,
385 			dst->sa_family);
386 		senderr(EAFNOSUPPORT);
387 	}
388 
389 #ifdef MPLS
390 	KERNEL_LOCK(1, NULL);
391 	{
392 		struct m_tag *mtag;
393 		mtag = m_tag_find(m, PACKET_TAG_MPLS, NULL);
394 		if (mtag != NULL) {
395 			/* Having the tag itself indicates it's MPLS */
396 			etype = htons(ETHERTYPE_MPLS);
397 			m_tag_delete(m, mtag);
398 		}
399 	}
400 	KERNEL_UNLOCK_ONE(NULL);
401 #endif
402 
403 	if (mcopy)
404 		(void)looutput(ifp, mcopy, dst, rt);
405 
406 	/* If no ether type is set, this must be a 802.2 formatted packet.
407 	 */
408 	if (etype == 0)
409 		etype = htons(m->m_pkthdr.len);
410 	/*
411 	 * Add local net header.  If no space in first mbuf,
412 	 * allocate another.
413 	 */
414 	M_PREPEND(m, sizeof (struct ether_header), M_DONTWAIT);
415 	if (m == 0)
416 		senderr(ENOBUFS);
417 	eh = mtod(m, struct ether_header *);
418 	/* Note: etype is already in network byte order. */
419 	(void)memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
420  	memcpy(eh->ether_dhost, edst, sizeof(edst));
421 	if (hdrcmplt)
422 		memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost));
423 	else
424 	 	memcpy(eh->ether_shost, CLLADDR(ifp->if_sadl),
425 		    sizeof(eh->ether_shost));
426 
427 #if NCARP > 0
428 	if (ifp0 != ifp && ifp0->if_type == IFT_CARP) {
429 	 	memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl),
430 		    sizeof(eh->ether_shost));
431 	}
432 #endif /* NCARP > 0 */
433 
434 	if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0)
435 		return (error);
436 	if (m == NULL)
437 		return (0);
438 
439 #if NBRIDGE > 0
440 	/*
441 	 * Bridges require special output handling.
442 	 */
443 	if (ifp->if_bridge)
444 		return (bridge_output(ifp, m, NULL, NULL));
445 #endif
446 
447 #if NCARP > 0
448 	if (ifp != ifp0)
449 		ifp0->if_obytes += m->m_pkthdr.len + ETHER_HDR_LEN;
450 #endif /* NCARP > 0 */
451 
452 #ifdef ALTQ
453 	KERNEL_LOCK(1, NULL);
454 	/*
455 	 * If ALTQ is enabled on the parent interface, do
456 	 * classification; the queueing discipline might not
457 	 * require classification, but might require the
458 	 * address family/header pointer in the pktattr.
459 	 */
460 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
461 		altq_etherclassify(&ifp->if_snd, m);
462 	KERNEL_UNLOCK_ONE(NULL);
463 #endif
464 	return ifq_enqueue(ifp, m);
465 
466 bad:
467 	if (m)
468 		m_freem(m);
469 	return (error);
470 }
471 
472 #ifdef ALTQ
473 /*
474  * This routine is a slight hack to allow a packet to be classified
475  * if the Ethernet headers are present.  It will go away when ALTQ's
476  * classification engine understands link headers.
477  */
478 void
479 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m)
480 {
481 	struct ether_header *eh;
482 	uint16_t ether_type;
483 	int hlen, af, hdrsize;
484 	void *hdr;
485 
486 	hlen = ETHER_HDR_LEN;
487 	eh = mtod(m, struct ether_header *);
488 
489 	ether_type = htons(eh->ether_type);
490 
491 	if (ether_type < ETHERMTU) {
492 		/* LLC/SNAP */
493 		struct llc *llc = (struct llc *)(eh + 1);
494 		hlen += 8;
495 
496 		if (m->m_len < hlen ||
497 		    llc->llc_dsap != LLC_SNAP_LSAP ||
498 		    llc->llc_ssap != LLC_SNAP_LSAP ||
499 		    llc->llc_control != LLC_UI) {
500 			/* Not SNAP. */
501 			goto bad;
502 		}
503 
504 		ether_type = htons(llc->llc_un.type_snap.ether_type);
505 	}
506 
507 	switch (ether_type) {
508 	case ETHERTYPE_IP:
509 		af = AF_INET;
510 		hdrsize = 20;		/* sizeof(struct ip) */
511 		break;
512 
513 	case ETHERTYPE_IPV6:
514 		af = AF_INET6;
515 		hdrsize = 40;		/* sizeof(struct ip6_hdr) */
516 		break;
517 
518 	default:
519 		af = AF_UNSPEC;
520 		hdrsize = 0;
521 		break;
522 	}
523 
524 	while (m->m_len <= hlen) {
525 		hlen -= m->m_len;
526 		m = m->m_next;
527 	}
528 	if (m->m_len < (hlen + hdrsize)) {
529 		/*
530 		 * protocol header not in a single mbuf.
531 		 * We can't cope with this situation right
532 		 * now (but it shouldn't ever happen, really, anyhow).
533 		 */
534 #ifdef DEBUG
535 		printf("altq_etherclassify: headers span multiple mbufs: "
536 		    "%d < %d\n", m->m_len, (hlen + hdrsize));
537 #endif
538 		goto bad;
539 	}
540 
541 	m->m_data += hlen;
542 	m->m_len -= hlen;
543 
544 	hdr = mtod(m, void *);
545 
546 	if (ALTQ_NEEDS_CLASSIFY(ifq))
547 		m->m_pkthdr.pattr_class =
548 		    (*ifq->altq_classify)(ifq->altq_clfier, m, af);
549 	m->m_pkthdr.pattr_af = af;
550 	m->m_pkthdr.pattr_hdr = hdr;
551 
552 	m->m_data -= hlen;
553 	m->m_len += hlen;
554 
555 	return;
556 
557  bad:
558 	m->m_pkthdr.pattr_class = NULL;
559 	m->m_pkthdr.pattr_hdr = NULL;
560 	m->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 
586 	if ((ifp->if_flags & IFF_UP) == 0) {
587 		m_freem(m);
588 		return;
589 	}
590 
591 #ifdef MBUFTRACE
592 	m_claimm(m, &ec->ec_rx_mowner);
593 #endif
594 	eh = mtod(m, struct ether_header *);
595 	etype = ntohs(eh->ether_type);
596 	ehlen = sizeof(*eh);
597 
598 	if(__predict_false(earlypkts < 100 || !rnd_initial_entropy)) {
599 		rnd_add_data(NULL, eh, ehlen, 0);
600 		earlypkts++;
601 	}
602 
603 	/*
604 	 * Determine if the packet is within its size limits.
605 	 */
606 	if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len >
607 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
608 		mutex_enter(&bigpktpps_lock);
609 		if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count,
610 			    bigpktppslim)) {
611 			printf("%s: discarding oversize frame (len=%d)\n",
612 			    ifp->if_xname, m->m_pkthdr.len);
613 		}
614 		mutex_exit(&bigpktpps_lock);
615 		m_freem(m);
616 		return;
617 	}
618 
619 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
620 		/*
621 		 * If this is not a simplex interface, drop the packet
622 		 * if it came from us.
623 		 */
624 		if ((ifp->if_flags & IFF_SIMPLEX) == 0 &&
625 		    memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost,
626 		    ETHER_ADDR_LEN) == 0) {
627 			m_freem(m);
628 			return;
629 		}
630 
631 		if (memcmp(etherbroadcastaddr,
632 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
633 			m->m_flags |= M_BCAST;
634 		else
635 			m->m_flags |= M_MCAST;
636 		ifp->if_imcasts++;
637 	}
638 
639 	/* If the CRC is still on the packet, trim it off. */
640 	if (m->m_flags & M_HASFCS) {
641 		m_adj(m, -ETHER_CRC_LEN);
642 		m->m_flags &= ~M_HASFCS;
643 	}
644 
645 	ifp->if_ibytes += m->m_pkthdr.len;
646 
647 #if NCARP > 0
648 	if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) {
649 		/*
650 		 * clear M_PROMISC, in case the packets comes from a
651 		 * vlan
652 		 */
653 		m->m_flags &= ~M_PROMISC;
654 		if (carp_input(m, (uint8_t *)&eh->ether_shost,
655 		    (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0)
656 			return;
657 	}
658 #endif /* NCARP > 0 */
659 	if ((m->m_flags & (M_BCAST|M_MCAST|M_PROMISC)) == 0 &&
660 	    (ifp->if_flags & IFF_PROMISC) != 0 &&
661 	    memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost,
662 		   ETHER_ADDR_LEN) != 0) {
663 		m->m_flags |= M_PROMISC;
664 	}
665 
666 	if ((m->m_flags & M_PROMISC) == 0) {
667 		if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
668 			return;
669 		if (m == NULL)
670 			return;
671 
672 		eh = mtod(m, struct ether_header *);
673 		etype = ntohs(eh->ether_type);
674 		ehlen = sizeof(*eh);
675 	}
676 
677 #if NAGR > 0
678 	if (ifp->if_agrprivate &&
679 	    __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) {
680 		m->m_flags &= ~M_PROMISC;
681 		agr_input(ifp, m);
682 		return;
683 	}
684 #endif /* NAGR > 0 */
685 
686 	/*
687 	 * If VLANs are configured on the interface, check to
688 	 * see if the device performed the decapsulation and
689 	 * provided us with the tag.
690 	 */
691 	if (ec->ec_nvlans && m_tag_find(m, PACKET_TAG_VLAN, NULL) != NULL) {
692 #if NVLAN > 0
693 		/*
694 		 * vlan_input() will either recursively call ether_input()
695 		 * or drop the packet.
696 		 */
697 		vlan_input(ifp, m);
698 #else
699 		m_freem(m);
700 #endif
701 		return;
702 	}
703 
704 	/*
705 	 * Handle protocols that expect to have the Ethernet header
706 	 * (and possibly FCS) intact.
707 	 */
708 	switch (etype) {
709 	case ETHERTYPE_VLAN: {
710 		struct ether_vlan_header *evl = (void *)eh;
711 		/*
712 		 * If there is a tag of 0, then the VLAN header was probably
713 		 * just being used to store the priority.  Extract the ether
714 		 * type, and if IP or IPV6, let them deal with it.
715 		 */
716 		if (m->m_len <= sizeof(*evl)
717 		    && EVL_VLANOFTAG(evl->evl_tag) == 0) {
718 			etype = ntohs(evl->evl_proto);
719 			ehlen = sizeof(*evl);
720 			if ((m->m_flags & M_PROMISC) == 0
721 			    && (etype == ETHERTYPE_IP
722 				|| etype == ETHERTYPE_IPV6))
723 				break;
724 		}
725 #if NVLAN > 0
726 		/*
727 		 * vlan_input() will either recursively call ether_input()
728 		 * or drop the packet.
729 		 */
730 		if (((struct ethercom *)ifp)->ec_nvlans != 0)
731 			vlan_input(ifp, m);
732 		else
733 #endif /* NVLAN > 0 */
734 			m_freem(m);
735 		return;
736 	}
737 #if NPPPOE > 0
738 	case ETHERTYPE_PPPOEDISC:
739 		pppoedisc_input(ifp, m);
740 		return;
741 	case ETHERTYPE_PPPOE:
742 		pppoe_input(ifp, m);
743 		return;
744 #endif /* NPPPOE > 0 */
745 	case ETHERTYPE_SLOWPROTOCOLS: {
746 		uint8_t subtype;
747 
748 #if defined(DIAGNOSTIC)
749 		if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype)) {
750 			panic("ether_input: too short slow protocol packet");
751 		}
752 #endif
753 		m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype);
754 		switch (subtype) {
755 #if NAGR > 0
756 		case SLOWPROTOCOLS_SUBTYPE_LACP:
757 			if (ifp->if_agrprivate) {
758 				ieee8023ad_lacp_input(ifp, m);
759 				return;
760 			}
761 			break;
762 
763 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
764 			if (ifp->if_agrprivate) {
765 				ieee8023ad_marker_input(ifp, m);
766 				return;
767 			}
768 			break;
769 #endif /* NAGR > 0 */
770 		default:
771 			if (subtype == 0 || subtype > 10) {
772 				/* illegal value */
773 				m_freem(m);
774 				return;
775 			}
776 			/* unknown subtype */
777 			break;
778 		}
779 		/* FALLTHROUGH */
780 	}
781 	default:
782 		if (m->m_flags & M_PROMISC) {
783 			m_freem(m);
784 			return;
785 		}
786 	}
787 
788 	/* If the CRC is still on the packet, trim it off. */
789 	if (m->m_flags & M_HASFCS) {
790 		m_adj(m, -ETHER_CRC_LEN);
791 		m->m_flags &= ~M_HASFCS;
792 	}
793 
794 	if (etype > ETHERMTU + sizeof (struct ether_header)) {
795 		/* Strip off the Ethernet header. */
796 		m_adj(m, ehlen);
797 
798 		switch (etype) {
799 #ifdef INET
800 		case ETHERTYPE_IP:
801 #ifdef GATEWAY
802 			if (ipflow_fastforward(m))
803 				return;
804 #endif
805 			pktq = ip_pktq;
806 			break;
807 
808 		case ETHERTYPE_ARP:
809 			isr = NETISR_ARP;
810 			inq = &arpintrq;
811 			break;
812 
813 		case ETHERTYPE_REVARP:
814 			revarpinput(m);	/* XXX queue? */
815 			return;
816 #endif
817 #ifdef INET6
818 		case ETHERTYPE_IPV6:
819 			if (__predict_false(!in6_present)) {
820 				m_freem(m);
821 				return;
822 			}
823 #ifdef GATEWAY
824 			if (ip6flow_fastforward(&m))
825 				return;
826 #endif
827 			pktq = ip6_pktq;
828 			break;
829 #endif
830 #ifdef NETATALK
831 		case ETHERTYPE_ATALK:
832 			isr = NETISR_ATALK;
833 			inq = &atintrq1;
834 			break;
835 		case ETHERTYPE_AARP:
836 			/* probably this should be done with a NETISR as well */
837 			aarpinput(ifp, m); /* XXX */
838 			return;
839 #endif /* NETATALK */
840 #ifdef MPLS
841 		case ETHERTYPE_MPLS:
842 			isr = NETISR_MPLS;
843 			inq = &mplsintrq;
844 			break;
845 #endif
846 		default:
847 			m_freem(m);
848 			return;
849 		}
850 	} else {
851 #if defined (LLC) || defined (NETATALK)
852 		l = (struct llc *)(eh+1);
853 		switch (l->llc_dsap) {
854 #ifdef NETATALK
855 		case LLC_SNAP_LSAP:
856 			switch (l->llc_control) {
857 			case LLC_UI:
858 				if (l->llc_ssap != LLC_SNAP_LSAP) {
859 					goto dropanyway;
860 				}
861 
862 				if (memcmp(&(l->llc_snap_org_code)[0],
863 				    at_org_code, sizeof(at_org_code)) == 0 &&
864 				    ntohs(l->llc_snap_ether_type) ==
865 				    ETHERTYPE_ATALK) {
866 					inq = &atintrq2;
867 					m_adj(m, sizeof(struct ether_header)
868 					    + sizeof(struct llc));
869 					isr = NETISR_ATALK;
870 					break;
871 				}
872 
873 				if (memcmp(&(l->llc_snap_org_code)[0],
874 				    aarp_org_code,
875 				    sizeof(aarp_org_code)) == 0 &&
876 				    ntohs(l->llc_snap_ether_type) ==
877 				    ETHERTYPE_AARP) {
878 					m_adj( m, sizeof(struct ether_header)
879 					    + sizeof(struct llc));
880 					aarpinput(ifp, m); /* XXX */
881 				    return;
882 				}
883 
884 			default:
885 				goto dropanyway;
886 			}
887 			break;
888 		dropanyway:
889 #endif
890 		default:
891 			m_freem(m);
892 			return;
893 		}
894 #else /* ISO || LLC || NETATALK*/
895 		m_freem(m);
896 		return;
897 #endif /* ISO || LLC || NETATALK*/
898 	}
899 
900 	if (__predict_true(pktq)) {
901 #ifdef NET_MPSAFE
902 		const u_int h = curcpu()->ci_index;
903 #else
904 		const uint32_t h = pktq_rps_hash(m);
905 #endif
906 		if (__predict_false(!pktq_enqueue(pktq, m, h))) {
907 			m_freem(m);
908 		}
909 		return;
910 	}
911 
912 	if (__predict_false(!inq)) {
913 		/* Should not happen. */
914 		m_freem(m);
915 		return;
916 	}
917 
918 	IFQ_LOCK(inq);
919 	if (IF_QFULL(inq)) {
920 		IF_DROP(inq);
921 		IFQ_UNLOCK(inq);
922 		m_freem(m);
923 	} else {
924 		IF_ENQUEUE(inq, m);
925 		IFQ_UNLOCK(inq);
926 		schednetisr(isr);
927 	}
928 }
929 
930 /*
931  * Convert Ethernet address to printable (loggable) representation.
932  */
933 char *
934 ether_sprintf(const u_char *ap)
935 {
936 	static char etherbuf[3 * ETHER_ADDR_LEN];
937 	return ether_snprintf(etherbuf, sizeof(etherbuf), ap);
938 }
939 
940 char *
941 ether_snprintf(char *buf, size_t len, const u_char *ap)
942 {
943 	char *cp = buf;
944 	size_t i;
945 
946 	for (i = 0; i < len / 3; i++) {
947 		*cp++ = hexdigits[*ap >> 4];
948 		*cp++ = hexdigits[*ap++ & 0xf];
949 		*cp++ = ':';
950 	}
951 	*--cp = '\0';
952 	return buf;
953 }
954 
955 /*
956  * Perform common duties while attaching to interface list
957  */
958 void
959 ether_ifattach(struct ifnet *ifp, const uint8_t *lla)
960 {
961 	struct ethercom *ec = (struct ethercom *)ifp;
962 
963 	ifp->if_extflags |= IFEF_OUTPUT_MPSAFE;
964 	ifp->if_type = IFT_ETHER;
965 	ifp->if_hdrlen = ETHER_HDR_LEN;
966 	ifp->if_dlt = DLT_EN10MB;
967 	ifp->if_mtu = ETHERMTU;
968 	ifp->if_output = ether_output;
969 	ifp->_if_input = ether_input;
970 	if (ifp->if_baudrate == 0)
971 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
972 
973 	if (lla != NULL)
974 		if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla));
975 
976 	LIST_INIT(&ec->ec_multiaddrs);
977 	ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
978 	ifp->if_broadcastaddr = etherbroadcastaddr;
979 	bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header));
980 #ifdef MBUFTRACE
981 	strlcpy(ec->ec_tx_mowner.mo_name, ifp->if_xname,
982 	    sizeof(ec->ec_tx_mowner.mo_name));
983 	strlcpy(ec->ec_tx_mowner.mo_descr, "tx",
984 	    sizeof(ec->ec_tx_mowner.mo_descr));
985 	strlcpy(ec->ec_rx_mowner.mo_name, ifp->if_xname,
986 	    sizeof(ec->ec_rx_mowner.mo_name));
987 	strlcpy(ec->ec_rx_mowner.mo_descr, "rx",
988 	    sizeof(ec->ec_rx_mowner.mo_descr));
989 	MOWNER_ATTACH(&ec->ec_tx_mowner);
990 	MOWNER_ATTACH(&ec->ec_rx_mowner);
991 	ifp->if_mowner = &ec->ec_tx_mowner;
992 #endif
993 }
994 
995 void
996 ether_ifdetach(struct ifnet *ifp)
997 {
998 	struct ethercom *ec = (void *) ifp;
999 	struct ether_multi *enm;
1000 
1001 	/*
1002 	 * Prevent further calls to ioctl (for example turning off
1003 	 * promiscuous mode from the bridge code), which eventually can
1004 	 * call if_init() which can cause panics because the interface
1005 	 * is in the process of being detached. Return device not configured
1006 	 * instead.
1007 	 */
1008 	ifp->if_ioctl = (int (*)(struct ifnet *, u_long, void *))enxio;
1009 
1010 #if NBRIDGE > 0
1011 	if (ifp->if_bridge)
1012 		bridge_ifdetach(ifp);
1013 #endif
1014 
1015 	bpf_detach(ifp);
1016 
1017 #if NVLAN > 0
1018 	if (ec->ec_nvlans)
1019 		vlan_ifdetach(ifp);
1020 #endif
1021 
1022 	mutex_enter(ec->ec_lock);
1023 	while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) {
1024 		LIST_REMOVE(enm, enm_list);
1025 		kmem_free(enm, sizeof(*enm));
1026 		ec->ec_multicnt--;
1027 	}
1028 	mutex_exit(ec->ec_lock);
1029 
1030 	mutex_destroy(ec->ec_lock);
1031 
1032 	ifp->if_mowner = NULL;
1033 	MOWNER_DETACH(&ec->ec_rx_mowner);
1034 	MOWNER_DETACH(&ec->ec_tx_mowner);
1035 }
1036 
1037 #if 0
1038 /*
1039  * This is for reference.  We have a table-driven version
1040  * of the little-endian crc32 generator, which is faster
1041  * than the double-loop.
1042  */
1043 uint32_t
1044 ether_crc32_le(const uint8_t *buf, size_t len)
1045 {
1046 	uint32_t c, crc, carry;
1047 	size_t i, j;
1048 
1049 	crc = 0xffffffffU;	/* initial value */
1050 
1051 	for (i = 0; i < len; i++) {
1052 		c = buf[i];
1053 		for (j = 0; j < 8; j++) {
1054 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1055 			crc >>= 1;
1056 			c >>= 1;
1057 			if (carry)
1058 				crc = (crc ^ ETHER_CRC_POLY_LE);
1059 		}
1060 	}
1061 
1062 	return (crc);
1063 }
1064 #else
1065 uint32_t
1066 ether_crc32_le(const uint8_t *buf, size_t len)
1067 {
1068 	static const uint32_t crctab[] = {
1069 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1070 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1071 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1072 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1073 	};
1074 	uint32_t crc;
1075 	size_t i;
1076 
1077 	crc = 0xffffffffU;	/* initial value */
1078 
1079 	for (i = 0; i < len; i++) {
1080 		crc ^= buf[i];
1081 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1082 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1083 	}
1084 
1085 	return (crc);
1086 }
1087 #endif
1088 
1089 uint32_t
1090 ether_crc32_be(const uint8_t *buf, size_t len)
1091 {
1092 	uint32_t c, crc, carry;
1093 	size_t i, j;
1094 
1095 	crc = 0xffffffffU;	/* initial value */
1096 
1097 	for (i = 0; i < len; i++) {
1098 		c = buf[i];
1099 		for (j = 0; j < 8; j++) {
1100 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1101 			crc <<= 1;
1102 			c >>= 1;
1103 			if (carry)
1104 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1105 		}
1106 	}
1107 
1108 	return (crc);
1109 }
1110 
1111 #ifdef INET
1112 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] =
1113     { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 };
1114 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] =
1115     { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff };
1116 #endif
1117 #ifdef INET6
1118 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] =
1119     { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 };
1120 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] =
1121     { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff };
1122 #endif
1123 
1124 /*
1125  * ether_aton implementation, not using a static buffer.
1126  */
1127 int
1128 ether_aton_r(u_char *dest, size_t len, const char *str)
1129 {
1130         const u_char *cp = (const void *)str;
1131 	u_char *ep;
1132 
1133 #define atox(c)	(((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10))
1134 
1135 	if (len < ETHER_ADDR_LEN)
1136 		return ENOSPC;
1137 
1138 	ep = dest + ETHER_ADDR_LEN;
1139 
1140 	while (*cp) {
1141                 if (!isxdigit(*cp))
1142                         return EINVAL;
1143 		*dest = atox(*cp);
1144 		cp++;
1145                 if (isxdigit(*cp)) {
1146                         *dest = (*dest << 4) | atox(*cp);
1147 			dest++;
1148 			cp++;
1149                 } else
1150 			dest++;
1151 		if (dest == ep)
1152 			return *cp == '\0' ? 0 : ENAMETOOLONG;
1153 		switch (*cp) {
1154 		case ':':
1155 		case '-':
1156 		case '.':
1157 			cp++;
1158 			break;
1159 		}
1160         }
1161 	return ENOBUFS;
1162 }
1163 
1164 /*
1165  * Convert a sockaddr into an Ethernet address or range of Ethernet
1166  * addresses.
1167  */
1168 int
1169 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN],
1170     uint8_t addrhi[ETHER_ADDR_LEN])
1171 {
1172 #ifdef INET
1173 	const struct sockaddr_in *sin;
1174 #endif /* INET */
1175 #ifdef INET6
1176 	const struct sockaddr_in6 *sin6;
1177 #endif /* INET6 */
1178 
1179 	switch (sa->sa_family) {
1180 
1181 	case AF_UNSPEC:
1182 		memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN);
1183 		memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1184 		break;
1185 
1186 #ifdef INET
1187 	case AF_INET:
1188 		sin = satocsin(sa);
1189 		if (sin->sin_addr.s_addr == INADDR_ANY) {
1190 			/*
1191 			 * An IP address of INADDR_ANY means listen to
1192 			 * or stop listening to all of the Ethernet
1193 			 * multicast addresses used for IP.
1194 			 * (This is for the sake of IP multicast routers.)
1195 			 */
1196 			memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN);
1197 			memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN);
1198 		}
1199 		else {
1200 			ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo);
1201 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1202 		}
1203 		break;
1204 #endif
1205 #ifdef INET6
1206 	case AF_INET6:
1207 		sin6 = satocsin6(sa);
1208 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1209 			/*
1210 			 * An IP6 address of 0 means listen to or stop
1211 			 * listening to all of the Ethernet multicast
1212 			 * address used for IP6.
1213 			 * (This is used for multicast routers.)
1214 			 */
1215 			memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN);
1216 			memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN);
1217 		} else {
1218 			ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo);
1219 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1220 		}
1221 		break;
1222 #endif
1223 
1224 	default:
1225 		return EAFNOSUPPORT;
1226 	}
1227 	return 0;
1228 }
1229 
1230 /*
1231  * Add an Ethernet multicast address or range of addresses to the list for a
1232  * given interface.
1233  */
1234 int
1235 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec)
1236 {
1237 	struct ether_multi *enm, *_enm;
1238 	u_char addrlo[ETHER_ADDR_LEN];
1239 	u_char addrhi[ETHER_ADDR_LEN];
1240 	int error = 0;
1241 
1242 	/* Allocate out of lock */
1243 	/* XXX still can be called in softint */
1244 	enm = kmem_intr_alloc(sizeof(*enm), KM_SLEEP);
1245 	if (enm == NULL)
1246 		return ENOBUFS;
1247 
1248 	mutex_enter(ec->ec_lock);
1249 	error = ether_multiaddr(sa, addrlo, addrhi);
1250 	if (error != 0)
1251 		goto out;
1252 
1253 	/*
1254 	 * Verify that we have valid Ethernet multicast addresses.
1255 	 */
1256 	if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) {
1257 		error = EINVAL;
1258 		goto out;
1259 	}
1260 	/*
1261 	 * See if the address range is already in the list.
1262 	 */
1263 	ETHER_LOOKUP_MULTI(addrlo, addrhi, ec, _enm);
1264 	if (_enm != NULL) {
1265 		/*
1266 		 * Found it; just increment the reference count.
1267 		 */
1268 		++_enm->enm_refcount;
1269 		error = 0;
1270 		goto out;
1271 	}
1272 	/*
1273 	 * Link a new multicast record into the interface's multicast list.
1274 	 */
1275 	memcpy(enm->enm_addrlo, addrlo, 6);
1276 	memcpy(enm->enm_addrhi, addrhi, 6);
1277 	enm->enm_refcount = 1;
1278 	LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list);
1279 	ec->ec_multicnt++;
1280 	/*
1281 	 * Return ENETRESET to inform the driver that the list has changed
1282 	 * and its reception filter should be adjusted accordingly.
1283 	 */
1284 	error = ENETRESET;
1285 	enm = NULL;
1286 out:
1287 	mutex_exit(ec->ec_lock);
1288 	if (enm != NULL)
1289 		kmem_free(enm, sizeof(*enm));
1290 	return error;
1291 }
1292 
1293 /*
1294  * Delete a multicast address record.
1295  */
1296 int
1297 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec)
1298 {
1299 	struct ether_multi *enm;
1300 	u_char addrlo[ETHER_ADDR_LEN];
1301 	u_char addrhi[ETHER_ADDR_LEN];
1302 	int error;
1303 
1304 	mutex_enter(ec->ec_lock);
1305 	error = ether_multiaddr(sa, addrlo, addrhi);
1306 	if (error != 0)
1307 		goto error;
1308 
1309 	/*
1310 	 * Look ur the address in our list.
1311 	 */
1312 	ETHER_LOOKUP_MULTI(addrlo, addrhi, ec, enm);
1313 	if (enm == NULL) {
1314 		error = ENXIO;
1315 		goto error;
1316 	}
1317 	if (--enm->enm_refcount != 0) {
1318 		/*
1319 		 * Still some claims to this record.
1320 		 */
1321 		error = 0;
1322 		goto error;
1323 	}
1324 	/*
1325 	 * No remaining claims to this record; unlink and free it.
1326 	 */
1327 	LIST_REMOVE(enm, enm_list);
1328 	ec->ec_multicnt--;
1329 	mutex_exit(ec->ec_lock);
1330 
1331 	kmem_free(enm, sizeof(*enm));
1332 	/*
1333 	 * Return ENETRESET to inform the driver that the list has changed
1334 	 * and its reception filter should be adjusted accordingly.
1335 	 */
1336 	return ENETRESET;
1337 error:
1338 	mutex_exit(ec->ec_lock);
1339 	return error;
1340 }
1341 
1342 void
1343 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb)
1344 {
1345 	ec->ec_ifflags_cb = cb;
1346 }
1347 
1348 /*
1349  * Common ioctls for Ethernet interfaces.  Note, we must be
1350  * called at splnet().
1351  */
1352 int
1353 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data)
1354 {
1355 	struct ethercom *ec = (void *) ifp;
1356 	struct eccapreq *eccr;
1357 	struct ifreq *ifr = (struct ifreq *)data;
1358 	struct if_laddrreq *iflr = data;
1359 	const struct sockaddr_dl *sdl;
1360 	static const uint8_t zero[ETHER_ADDR_LEN];
1361 	int error;
1362 
1363 	switch (cmd) {
1364 	case SIOCINITIFADDR:
1365 	    {
1366 		struct ifaddr *ifa = (struct ifaddr *)data;
1367 		if (ifa->ifa_addr->sa_family != AF_LINK
1368 		    && (ifp->if_flags & (IFF_UP|IFF_RUNNING)) !=
1369 		       (IFF_UP|IFF_RUNNING)) {
1370 			ifp->if_flags |= IFF_UP;
1371 			if ((error = (*ifp->if_init)(ifp)) != 0)
1372 				return error;
1373 		}
1374 #ifdef INET
1375 		if (ifa->ifa_addr->sa_family == AF_INET)
1376 			arp_ifinit(ifp, ifa);
1377 #endif /* INET */
1378 		return 0;
1379 	    }
1380 
1381 	case SIOCSIFMTU:
1382 	    {
1383 		int maxmtu;
1384 
1385 		if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU)
1386 			maxmtu = ETHERMTU_JUMBO;
1387 		else
1388 			maxmtu = ETHERMTU;
1389 
1390 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu)
1391 			return EINVAL;
1392 		else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET)
1393 			return error;
1394 		else if (ifp->if_flags & IFF_UP) {
1395 			/* Make sure the device notices the MTU change. */
1396 			return (*ifp->if_init)(ifp);
1397 		} else
1398 			return 0;
1399 	    }
1400 
1401 	case SIOCSIFFLAGS:
1402 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
1403 			return error;
1404 		switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
1405 		case IFF_RUNNING:
1406 			/*
1407 			 * If interface is marked down and it is running,
1408 			 * then stop and disable it.
1409 			 */
1410 			(*ifp->if_stop)(ifp, 1);
1411 			break;
1412 		case IFF_UP:
1413 			/*
1414 			 * If interface is marked up and it is stopped, then
1415 			 * start it.
1416 			 */
1417 			return (*ifp->if_init)(ifp);
1418 		case IFF_UP|IFF_RUNNING:
1419 			error = 0;
1420 			if (ec->ec_ifflags_cb == NULL ||
1421 			    (error = (*ec->ec_ifflags_cb)(ec)) == ENETRESET) {
1422 				/*
1423 				 * Reset the interface to pick up
1424 				 * changes in any other flags that
1425 				 * affect the hardware state.
1426 				 */
1427 				return (*ifp->if_init)(ifp);
1428 			} else
1429 				return error;
1430 		case 0:
1431 			break;
1432 		}
1433 		return 0;
1434 	case SIOCGETHERCAP:
1435 		eccr = (struct eccapreq *)data;
1436 		eccr->eccr_capabilities = ec->ec_capabilities;
1437 		eccr->eccr_capenable = ec->ec_capenable;
1438 		return 0;
1439 	case SIOCADDMULTI:
1440 		return ether_addmulti(ifreq_getaddr(cmd, ifr), ec);
1441 	case SIOCDELMULTI:
1442 		return ether_delmulti(ifreq_getaddr(cmd, ifr), ec);
1443 	case SIOCSIFMEDIA:
1444 	case SIOCGIFMEDIA:
1445 		if (ec->ec_mii == NULL)
1446 			return ENOTTY;
1447 		return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media, cmd);
1448 	case SIOCALIFADDR:
1449 		sdl = satocsdl(sstocsa(&iflr->addr));
1450 		if (sdl->sdl_family != AF_LINK)
1451 			;
1452 		else if (ETHER_IS_MULTICAST(CLLADDR(sdl)))
1453 			return EINVAL;
1454 		else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0)
1455 			return EINVAL;
1456 		/*FALLTHROUGH*/
1457 	default:
1458 		return ifioctl_common(ifp, cmd, data);
1459 	}
1460 	return 0;
1461 }
1462 
1463 /*
1464  * Enable/disable passing VLAN packets if the parent interface supports it.
1465  * Return:
1466  * 	 0: Ok
1467  *	-1: Parent interface does not support vlans
1468  *	>0: Error
1469  */
1470 int
1471 ether_enable_vlan_mtu(struct ifnet *ifp)
1472 {
1473 	int error;
1474 	struct ethercom *ec = (void *)ifp;
1475 
1476 	/* Parent does not support VLAN's */
1477 	if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0)
1478 		return -1;
1479 
1480 	/*
1481 	 * Parent supports the VLAN_MTU capability,
1482 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames;
1483 	 * enable it.
1484 	 */
1485 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1486 
1487 	/* Interface is down, defer for later */
1488 	if ((ifp->if_flags & IFF_UP) == 0)
1489 		return 0;
1490 
1491 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1492 		return 0;
1493 
1494 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1495 	return error;
1496 }
1497 
1498 int
1499 ether_disable_vlan_mtu(struct ifnet *ifp)
1500 {
1501 	int error;
1502 	struct ethercom *ec = (void *)ifp;
1503 
1504 	/* We still have VLAN's, defer for later */
1505 	if (ec->ec_nvlans != 0)
1506 		return 0;
1507 
1508 	/* Parent does not support VLAB's, nothing to do. */
1509 	if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0)
1510 		return -1;
1511 
1512 	/*
1513 	 * Disable Tx/Rx of VLAN-sized frames.
1514 	 */
1515 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1516 
1517 	/* Interface is down, defer for later */
1518 	if ((ifp->if_flags & IFF_UP) == 0)
1519 		return 0;
1520 
1521 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1522 		return 0;
1523 
1524 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1525 	return error;
1526 }
1527 
1528 static int
1529 ether_multicast_sysctl(SYSCTLFN_ARGS)
1530 {
1531 	struct ether_multi *enm;
1532 	struct ifnet *ifp;
1533 	struct ethercom *ec;
1534 	int error = 0;
1535 	size_t written;
1536 	struct psref psref;
1537 	int bound;
1538 	unsigned int multicnt;
1539 	struct ether_multi_sysctl *addrs;
1540 	int i;
1541 
1542 	if (namelen != 1)
1543 		return EINVAL;
1544 
1545 	bound = curlwp_bind();
1546 	ifp = if_get_byindex(name[0], &psref);
1547 	if (ifp == NULL) {
1548 		error = ENODEV;
1549 		goto out;
1550 	}
1551 	if (ifp->if_type != IFT_ETHER) {
1552 		if_put(ifp, &psref);
1553 		*oldlenp = 0;
1554 		goto out;
1555 	}
1556 	ec = (struct ethercom *)ifp;
1557 
1558 	if (oldp == NULL) {
1559 		if_put(ifp, &psref);
1560 		*oldlenp = ec->ec_multicnt * sizeof(*addrs);
1561 		goto out;
1562 	}
1563 
1564 	/*
1565 	 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which
1566 	 * is sleepable, with holding it. Copy data to a local buffer first
1567 	 * with holding it and then call sysctl_copyout without holding it.
1568 	 */
1569 retry:
1570 	multicnt = ec->ec_multicnt;
1571 	addrs = kmem_alloc(sizeof(*addrs) * multicnt, KM_SLEEP);
1572 
1573 	mutex_enter(ec->ec_lock);
1574 	if (multicnt < ec->ec_multicnt) {
1575 		/* The number of multicast addresses have increased */
1576 		mutex_exit(ec->ec_lock);
1577 		kmem_free(addrs, sizeof(*addrs) * multicnt);
1578 		goto retry;
1579 	}
1580 
1581 	i = 0;
1582 	LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) {
1583 		struct ether_multi_sysctl *addr = &addrs[i];
1584 		addr->enm_refcount = enm->enm_refcount;
1585 		memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN);
1586 		memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN);
1587 		i++;
1588 	}
1589 	mutex_exit(ec->ec_lock);
1590 
1591 	error = 0;
1592 	written = 0;
1593 	for (i = 0; i < multicnt; i++) {
1594 		struct ether_multi_sysctl *addr = &addrs[i];
1595 
1596 		if (written + sizeof(*addr) > *oldlenp)
1597 			break;
1598 		error = sysctl_copyout(l, addr, oldp, sizeof(*addr));
1599 		if (error)
1600 			break;
1601 		written += sizeof(*addr);
1602 		oldp = (char *)oldp + sizeof(*addr);
1603 	}
1604 	kmem_free(addrs, sizeof(*addrs) * multicnt);
1605 
1606 	if_put(ifp, &psref);
1607 
1608 	*oldlenp = written;
1609 out:
1610 	curlwp_bindx(bound);
1611 	return error;
1612 }
1613 
1614 static void
1615 ether_sysctl_setup(struct sysctllog **clog)
1616 {
1617 	const struct sysctlnode *rnode = NULL;
1618 
1619 	sysctl_createv(clog, 0, NULL, &rnode,
1620 		       CTLFLAG_PERMANENT,
1621 		       CTLTYPE_NODE, "ether",
1622 		       SYSCTL_DESCR("Ethernet-specific information"),
1623 		       NULL, 0, NULL, 0,
1624 		       CTL_NET, CTL_CREATE, CTL_EOL);
1625 
1626 	sysctl_createv(clog, 0, &rnode, NULL,
1627 		       CTLFLAG_PERMANENT,
1628 		       CTLTYPE_NODE, "multicast",
1629 		       SYSCTL_DESCR("multicast addresses"),
1630 		       ether_multicast_sysctl, 0, NULL, 0,
1631 		       CTL_CREATE, CTL_EOL);
1632 }
1633 
1634 void
1635 etherinit(void)
1636 {
1637 
1638 	mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET);
1639 	ether_sysctl_setup(NULL);
1640 }
1641