xref: /netbsd-src/sys/net/if_ethersubr.c (revision 82d56013d7b633d116a93943de88e08335357a7c)
1 /*	$NetBSD: if_ethersubr.c,v 1.293 2021/05/17 04:07:43 yamaguchi Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (c) 1982, 1989, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)if_ethersubr.c	8.2 (Berkeley) 4/4/96
61  */
62 
63 #include <sys/cdefs.h>
64 __KERNEL_RCSID(0, "$NetBSD: if_ethersubr.c,v 1.293 2021/05/17 04:07:43 yamaguchi Exp $");
65 
66 #ifdef _KERNEL_OPT
67 #include "opt_inet.h"
68 #include "opt_atalk.h"
69 #include "opt_mbuftrace.h"
70 #include "opt_mpls.h"
71 #include "opt_gateway.h"
72 #include "opt_pppoe.h"
73 #include "opt_net_mpsafe.h"
74 #endif
75 
76 #include "vlan.h"
77 #include "pppoe.h"
78 #include "bridge.h"
79 #include "arp.h"
80 #include "agr.h"
81 #include "lagg.h"
82 
83 #include <sys/sysctl.h>
84 #include <sys/mbuf.h>
85 #include <sys/mutex.h>
86 #include <sys/ioctl.h>
87 #include <sys/errno.h>
88 #include <sys/device.h>
89 #include <sys/entropy.h>
90 #include <sys/rndsource.h>
91 #include <sys/cpu.h>
92 #include <sys/kmem.h>
93 
94 #include <net/if.h>
95 #include <net/netisr.h>
96 #include <net/route.h>
97 #include <net/if_llc.h>
98 #include <net/if_dl.h>
99 #include <net/if_types.h>
100 #include <net/pktqueue.h>
101 
102 #include <net/if_media.h>
103 #include <dev/mii/mii.h>
104 #include <dev/mii/miivar.h>
105 
106 #if NARP == 0
107 /*
108  * XXX there should really be a way to issue this warning from within config(8)
109  */
110 #error You have included NETATALK or a pseudo-device in your configuration that depends on the presence of ethernet interfaces, but have no such interfaces configured. Check if you really need pseudo-device bridge, pppoe, vlan or options NETATALK.
111 #endif
112 
113 #include <net/bpf.h>
114 
115 #include <net/if_ether.h>
116 #include <net/if_vlanvar.h>
117 
118 #if NPPPOE > 0
119 #include <net/if_pppoe.h>
120 #endif
121 
122 #if NAGR > 0
123 #include <net/agr/ieee8023_slowprotocols.h>	/* XXX */
124 #include <net/agr/ieee8023ad.h>
125 #include <net/agr/if_agrvar.h>
126 #endif
127 
128 #include <net/lagg/if_laggvar.h>
129 
130 #if NBRIDGE > 0
131 #include <net/if_bridgevar.h>
132 #endif
133 
134 #include <netinet/in.h>
135 #ifdef INET
136 #include <netinet/in_var.h>
137 #endif
138 #include <netinet/if_inarp.h>
139 
140 #ifdef INET6
141 #ifndef INET
142 #include <netinet/in.h>
143 #endif
144 #include <netinet6/in6_var.h>
145 #include <netinet6/nd6.h>
146 #endif
147 
148 #include "carp.h"
149 #if NCARP > 0
150 #include <netinet/ip_carp.h>
151 #endif
152 
153 #ifdef NETATALK
154 #include <netatalk/at.h>
155 #include <netatalk/at_var.h>
156 #include <netatalk/at_extern.h>
157 
158 #define llc_snap_org_code llc_un.type_snap.org_code
159 #define llc_snap_ether_type llc_un.type_snap.ether_type
160 
161 extern u_char	at_org_code[3];
162 extern u_char	aarp_org_code[3];
163 #endif /* NETATALK */
164 
165 #ifdef MPLS
166 #include <netmpls/mpls.h>
167 #include <netmpls/mpls_var.h>
168 #endif
169 
170 CTASSERT(sizeof(struct ether_addr) == 6);
171 CTASSERT(sizeof(struct ether_header) == 14);
172 
173 #ifdef DIAGNOSTIC
174 static struct timeval bigpktppslim_last;
175 static int bigpktppslim = 2;	/* XXX */
176 static int bigpktpps_count;
177 static kmutex_t bigpktpps_lock __cacheline_aligned;
178 #endif
179 
180 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] =
181     { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
182 const uint8_t ethermulticastaddr_slowprotocols[ETHER_ADDR_LEN] =
183     { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x02 };
184 #define senderr(e) { error = (e); goto bad;}
185 
186 /* if_lagg(4) support */
187 struct mbuf *(*lagg_input_ethernet_p)(struct ifnet *, struct mbuf *);
188 
189 static int ether_output(struct ifnet *, struct mbuf *,
190     const struct sockaddr *, const struct rtentry *);
191 
192 /*
193  * Ethernet output routine.
194  * Encapsulate a packet of type family for the local net.
195  * Assumes that ifp is actually pointer to ethercom structure.
196  */
197 static int
198 ether_output(struct ifnet * const ifp0, struct mbuf * const m0,
199     const struct sockaddr * const dst, const struct rtentry *rt)
200 {
201 	uint8_t esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
202 	uint16_t etype = 0;
203 	int error = 0, hdrcmplt = 0;
204 	struct mbuf *m = m0;
205 	struct mbuf *mcopy = NULL;
206 	struct ether_header *eh;
207 	struct ifnet *ifp = ifp0;
208 #ifdef INET
209 	struct arphdr *ah;
210 #endif
211 #ifdef NETATALK
212 	struct at_ifaddr *aa;
213 #endif
214 
215 #ifdef MBUFTRACE
216 	m_claimm(m, ifp->if_mowner);
217 #endif
218 
219 #if NCARP > 0
220 	if (ifp->if_type == IFT_CARP) {
221 		struct ifaddr *ifa;
222 		int s = pserialize_read_enter();
223 
224 		/* loop back if this is going to the carp interface */
225 		if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP &&
226 		    (ifa = ifa_ifwithaddr(dst)) != NULL) {
227 			if (ifa->ifa_ifp == ifp0) {
228 				pserialize_read_exit(s);
229 				return looutput(ifp0, m, dst, rt);
230 			}
231 		}
232 		pserialize_read_exit(s);
233 
234 		ifp = ifp->if_carpdev;
235 		/* ac = (struct arpcom *)ifp; */
236 
237 		if ((ifp0->if_flags & (IFF_UP | IFF_RUNNING)) !=
238 		    (IFF_UP | IFF_RUNNING))
239 			senderr(ENETDOWN);
240 	}
241 #endif
242 
243 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
244 		senderr(ENETDOWN);
245 
246 	switch (dst->sa_family) {
247 
248 #ifdef INET
249 	case AF_INET:
250 		if (m->m_flags & M_BCAST) {
251 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
252 		} else if (m->m_flags & M_MCAST) {
253 			ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst);
254 		} else {
255 			error = arpresolve(ifp0, rt, m, dst, edst, sizeof(edst));
256 			if (error)
257 				return (error == EWOULDBLOCK) ? 0 : error;
258 		}
259 		/* If broadcasting on a simplex interface, loopback a copy */
260 		if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
261 			mcopy = m_copypacket(m, M_DONTWAIT);
262 		etype = htons(ETHERTYPE_IP);
263 		break;
264 
265 	case AF_ARP:
266 		ah = mtod(m, struct arphdr *);
267 		if (m->m_flags & M_BCAST) {
268 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
269 		} else {
270 			void *tha = ar_tha(ah);
271 
272 			if (tha == NULL) {
273 				/* fake with ARPHRD_IEEE1394 */
274 				m_freem(m);
275 				return 0;
276 			}
277 			memcpy(edst, tha, sizeof(edst));
278 		}
279 
280 		ah->ar_hrd = htons(ARPHRD_ETHER);
281 
282 		switch (ntohs(ah->ar_op)) {
283 		case ARPOP_REVREQUEST:
284 		case ARPOP_REVREPLY:
285 			etype = htons(ETHERTYPE_REVARP);
286 			break;
287 
288 		case ARPOP_REQUEST:
289 		case ARPOP_REPLY:
290 		default:
291 			etype = htons(ETHERTYPE_ARP);
292 		}
293 		break;
294 #endif
295 
296 #ifdef INET6
297 	case AF_INET6:
298 		if (m->m_flags & M_BCAST) {
299 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
300 		} else if (m->m_flags & M_MCAST) {
301 			ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr,
302 			    edst);
303 		} else {
304 			error = nd6_resolve(ifp0, rt, m, dst, edst,
305 			    sizeof(edst));
306 			if (error)
307 				return (error == EWOULDBLOCK) ? 0 : error;
308 		}
309 		etype = htons(ETHERTYPE_IPV6);
310 		break;
311 #endif
312 
313 #ifdef NETATALK
314 	case AF_APPLETALK: {
315 		struct ifaddr *ifa;
316 		int s;
317 
318 		KERNEL_LOCK(1, NULL);
319 
320 		if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) {
321 			KERNEL_UNLOCK_ONE(NULL);
322 			return 0;
323 		}
324 
325 		/*
326 		 * ifaddr is the first thing in at_ifaddr
327 		 */
328 		s = pserialize_read_enter();
329 		ifa = at_ifawithnet((const struct sockaddr_at *)dst, ifp);
330 		if (ifa == NULL) {
331 			pserialize_read_exit(s);
332 			KERNEL_UNLOCK_ONE(NULL);
333 			senderr(EADDRNOTAVAIL);
334 		}
335 		aa = (struct at_ifaddr *)ifa;
336 
337 		/*
338 		 * In the phase 2 case, we need to prepend an mbuf for the
339 		 * llc header.
340 		 */
341 		if (aa->aa_flags & AFA_PHASE2) {
342 			struct llc llc;
343 
344 			M_PREPEND(m, sizeof(struct llc), M_DONTWAIT);
345 			if (m == NULL) {
346 				pserialize_read_exit(s);
347 				KERNEL_UNLOCK_ONE(NULL);
348 				senderr(ENOBUFS);
349 			}
350 
351 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
352 			llc.llc_control = LLC_UI;
353 			memcpy(llc.llc_snap_org_code, at_org_code,
354 			    sizeof(llc.llc_snap_org_code));
355 			llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK);
356 			memcpy(mtod(m, void *), &llc, sizeof(struct llc));
357 		} else {
358 			etype = htons(ETHERTYPE_ATALK);
359 		}
360 		pserialize_read_exit(s);
361 		KERNEL_UNLOCK_ONE(NULL);
362 		break;
363 	}
364 #endif /* NETATALK */
365 
366 	case pseudo_AF_HDRCMPLT:
367 		hdrcmplt = 1;
368 		memcpy(esrc,
369 		    ((const struct ether_header *)dst->sa_data)->ether_shost,
370 		    sizeof(esrc));
371 		/* FALLTHROUGH */
372 
373 	case AF_UNSPEC:
374 		memcpy(edst,
375 		    ((const struct ether_header *)dst->sa_data)->ether_dhost,
376 		    sizeof(edst));
377 		/* AF_UNSPEC doesn't swap the byte order of the ether_type. */
378 		etype = ((const struct ether_header *)dst->sa_data)->ether_type;
379 		break;
380 
381 	default:
382 		printf("%s: can't handle af%d\n", ifp->if_xname,
383 		    dst->sa_family);
384 		senderr(EAFNOSUPPORT);
385 	}
386 
387 #ifdef MPLS
388 	{
389 		struct m_tag *mtag;
390 		mtag = m_tag_find(m, PACKET_TAG_MPLS);
391 		if (mtag != NULL) {
392 			/* Having the tag itself indicates it's MPLS */
393 			etype = htons(ETHERTYPE_MPLS);
394 			m_tag_delete(m, mtag);
395 		}
396 	}
397 #endif
398 
399 	if (mcopy)
400 		(void)looutput(ifp, mcopy, dst, rt);
401 
402 	KASSERT((m->m_flags & M_PKTHDR) != 0);
403 
404 	/*
405 	 * If no ether type is set, this must be a 802.2 formatted packet.
406 	 */
407 	if (etype == 0)
408 		etype = htons(m->m_pkthdr.len);
409 
410 	/*
411 	 * Add local net header. If no space in first mbuf, allocate another.
412 	 */
413 	M_PREPEND(m, sizeof(struct ether_header), M_DONTWAIT);
414 	if (m == NULL)
415 		senderr(ENOBUFS);
416 
417 	eh = mtod(m, struct ether_header *);
418 	/* Note: etype is already in network byte order. */
419 	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 
428 #if NCARP > 0
429 	if (ifp0 != ifp && ifp0->if_type == IFT_CARP) {
430 	 	memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl),
431 		    sizeof(eh->ether_shost));
432 	}
433 #endif
434 
435 	if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0)
436 		return error;
437 	if (m == NULL)
438 		return 0;
439 
440 #if NBRIDGE > 0
441 	/*
442 	 * Bridges require special output handling.
443 	 */
444 	if (ifp->if_bridge)
445 		return bridge_output(ifp, m, NULL, NULL);
446 #endif
447 
448 #if NCARP > 0
449 	if (ifp != ifp0)
450 		if_statadd(ifp0, if_obytes, m->m_pkthdr.len + ETHER_HDR_LEN);
451 #endif
452 
453 #ifdef ALTQ
454 	KERNEL_LOCK(1, NULL);
455 	/*
456 	 * If ALTQ is enabled on the parent interface, do
457 	 * classification; the queueing discipline might not
458 	 * require classification, but might require the
459 	 * address family/header pointer in the pktattr.
460 	 */
461 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
462 		altq_etherclassify(&ifp->if_snd, m);
463 	KERNEL_UNLOCK_ONE(NULL);
464 #endif
465 	return ifq_enqueue(ifp, m);
466 
467 bad:
468 	if_statinc(ifp, if_oerrors);
469 	if (m)
470 		m_freem(m);
471 	return error;
472 }
473 
474 #ifdef ALTQ
475 /*
476  * This routine is a slight hack to allow a packet to be classified
477  * if the Ethernet headers are present.  It will go away when ALTQ's
478  * classification engine understands link headers.
479  *
480  * XXX: We may need to do m_pullups here. First to ensure struct ether_header
481  * is indeed contiguous, then to read the LLC and so on.
482  */
483 void
484 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m)
485 {
486 	struct ether_header *eh;
487 	struct mbuf *mtop = m;
488 	uint16_t ether_type;
489 	int hlen, af, hdrsize;
490 	void *hdr;
491 
492 	KASSERT((mtop->m_flags & M_PKTHDR) != 0);
493 
494 	hlen = ETHER_HDR_LEN;
495 	eh = mtod(m, struct ether_header *);
496 
497 	ether_type = htons(eh->ether_type);
498 
499 	if (ether_type < ETHERMTU) {
500 		/* LLC/SNAP */
501 		struct llc *llc = (struct llc *)(eh + 1);
502 		hlen += 8;
503 
504 		if (m->m_len < hlen ||
505 		    llc->llc_dsap != LLC_SNAP_LSAP ||
506 		    llc->llc_ssap != LLC_SNAP_LSAP ||
507 		    llc->llc_control != LLC_UI) {
508 			/* Not SNAP. */
509 			goto bad;
510 		}
511 
512 		ether_type = htons(llc->llc_un.type_snap.ether_type);
513 	}
514 
515 	switch (ether_type) {
516 	case ETHERTYPE_IP:
517 		af = AF_INET;
518 		hdrsize = 20;		/* sizeof(struct ip) */
519 		break;
520 
521 	case ETHERTYPE_IPV6:
522 		af = AF_INET6;
523 		hdrsize = 40;		/* sizeof(struct ip6_hdr) */
524 		break;
525 
526 	default:
527 		af = AF_UNSPEC;
528 		hdrsize = 0;
529 		break;
530 	}
531 
532 	while (m->m_len <= hlen) {
533 		hlen -= m->m_len;
534 		m = m->m_next;
535 		if (m == NULL)
536 			goto bad;
537 	}
538 
539 	if (m->m_len < (hlen + hdrsize)) {
540 		/*
541 		 * protocol header not in a single mbuf.
542 		 * We can't cope with this situation right
543 		 * now (but it shouldn't ever happen, really, anyhow).
544 		 */
545 #ifdef DEBUG
546 		printf("altq_etherclassify: headers span multiple mbufs: "
547 		    "%d < %d\n", m->m_len, (hlen + hdrsize));
548 #endif
549 		goto bad;
550 	}
551 
552 	m->m_data += hlen;
553 	m->m_len -= hlen;
554 
555 	hdr = mtod(m, void *);
556 
557 	if (ALTQ_NEEDS_CLASSIFY(ifq)) {
558 		mtop->m_pkthdr.pattr_class =
559 		    (*ifq->altq_classify)(ifq->altq_clfier, m, af);
560 	}
561 	mtop->m_pkthdr.pattr_af = af;
562 	mtop->m_pkthdr.pattr_hdr = hdr;
563 
564 	m->m_data -= hlen;
565 	m->m_len += hlen;
566 
567 	return;
568 
569 bad:
570 	mtop->m_pkthdr.pattr_class = NULL;
571 	mtop->m_pkthdr.pattr_hdr = NULL;
572 	mtop->m_pkthdr.pattr_af = AF_UNSPEC;
573 }
574 #endif /* ALTQ */
575 
576 #if defined (LLC) || defined (NETATALK)
577 static void
578 ether_input_llc(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh)
579 {
580 	struct ifqueue *inq = NULL;
581 	int isr = 0;
582 	struct llc *l;
583 
584 	if (m->m_len < sizeof(*eh) + sizeof(struct llc))
585 		goto drop;
586 
587 	l = (struct llc *)(eh+1);
588 	switch (l->llc_dsap) {
589 #ifdef NETATALK
590 	case LLC_SNAP_LSAP:
591 		switch (l->llc_control) {
592 		case LLC_UI:
593 			if (l->llc_ssap != LLC_SNAP_LSAP)
594 				goto drop;
595 
596 			if (memcmp(&(l->llc_snap_org_code)[0],
597 			    at_org_code, sizeof(at_org_code)) == 0 &&
598 			    ntohs(l->llc_snap_ether_type) ==
599 			    ETHERTYPE_ATALK) {
600 				inq = &atintrq2;
601 				m_adj(m, sizeof(struct ether_header)
602 				    + sizeof(struct llc));
603 				isr = NETISR_ATALK;
604 				break;
605 			}
606 
607 			if (memcmp(&(l->llc_snap_org_code)[0],
608 			    aarp_org_code,
609 			    sizeof(aarp_org_code)) == 0 &&
610 			    ntohs(l->llc_snap_ether_type) ==
611 			    ETHERTYPE_AARP) {
612 				m_adj(m, sizeof(struct ether_header)
613 				    + sizeof(struct llc));
614 				aarpinput(ifp, m); /* XXX queue? */
615 				return;
616 			}
617 
618 		default:
619 			goto drop;
620 		}
621 		break;
622 #endif
623 	default:
624 		goto drop;
625 	}
626 
627 	KASSERT(inq != NULL);
628 	IFQ_ENQUEUE_ISR(inq, m, isr);
629 	return;
630 
631 drop:
632 	m_freem(m);
633 	if_statinc(ifp, if_ierrors); /* XXX should have a dedicated counter? */
634 	return;
635 }
636 #endif /* defined (LLC) || defined (NETATALK) */
637 
638 /*
639  * Process a received Ethernet packet;
640  * the packet is in the mbuf chain m with
641  * the ether header.
642  */
643 void
644 ether_input(struct ifnet *ifp, struct mbuf *m)
645 {
646 	struct ethercom *ec = (struct ethercom *) ifp;
647 	pktqueue_t *pktq = NULL;
648 	struct ifqueue *inq = NULL;
649 	uint16_t etype;
650 	struct ether_header *eh;
651 	size_t ehlen;
652 	static int earlypkts;
653 	int isr = 0;
654 
655 	KASSERT(!cpu_intr_p());
656 	KASSERT((m->m_flags & M_PKTHDR) != 0);
657 
658 	if ((ifp->if_flags & IFF_UP) == 0)
659 		goto drop;
660 
661 #ifdef MBUFTRACE
662 	m_claimm(m, &ec->ec_rx_mowner);
663 #endif
664 
665 	if (__predict_false(m->m_len < sizeof(*eh))) {
666 		if ((m = m_pullup(m, sizeof(*eh))) == NULL)
667 			goto dropped;
668 	}
669 
670 	eh = mtod(m, struct ether_header *);
671 	etype = ntohs(eh->ether_type);
672 	ehlen = sizeof(*eh);
673 
674 	if (__predict_false(earlypkts < 100 ||
675 		entropy_epoch() == (unsigned)-1)) {
676 		rnd_add_data(NULL, eh, ehlen, 0);
677 		earlypkts++;
678 	}
679 
680 	/*
681 	 * Determine if the packet is within its size limits. For MPLS the
682 	 * header length is variable, so we skip the check.
683 	 */
684 	if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len >
685 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
686 #ifdef DIAGNOSTIC
687 		mutex_enter(&bigpktpps_lock);
688 		if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count,
689 		    bigpktppslim)) {
690 			printf("%s: discarding oversize frame (len=%d)\n",
691 			    ifp->if_xname, m->m_pkthdr.len);
692 		}
693 		mutex_exit(&bigpktpps_lock);
694 #endif
695 		goto drop;
696 	}
697 
698 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
699 		/*
700 		 * If this is not a simplex interface, drop the packet
701 		 * if it came from us.
702 		 */
703 		if ((ifp->if_flags & IFF_SIMPLEX) == 0 &&
704 		    memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost,
705 		    ETHER_ADDR_LEN) == 0) {
706 			goto drop;
707 		}
708 
709 		if (memcmp(etherbroadcastaddr,
710 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
711 			m->m_flags |= M_BCAST;
712 		else
713 			m->m_flags |= M_MCAST;
714 		if_statinc(ifp, if_imcasts);
715 	}
716 
717 	/* If the CRC is still on the packet, trim it off. */
718 	if (m->m_flags & M_HASFCS) {
719 		m_adj(m, -ETHER_CRC_LEN);
720 		m->m_flags &= ~M_HASFCS;
721 	}
722 
723 	if_statadd(ifp, if_ibytes, m->m_pkthdr.len);
724 
725 #if NCARP > 0
726 	if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) {
727 		/*
728 		 * Clear M_PROMISC, in case the packet comes from a
729 		 * vlan.
730 		 */
731 		m->m_flags &= ~M_PROMISC;
732 		if (carp_input(m, (uint8_t *)&eh->ether_shost,
733 		    (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0)
734 			return;
735 	}
736 #endif
737 
738 	if ((m->m_flags & (M_BCAST | M_MCAST | M_PROMISC)) == 0 &&
739 	    (ifp->if_flags & IFF_PROMISC) != 0 &&
740 	    memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost,
741 	     ETHER_ADDR_LEN) != 0) {
742 		m->m_flags |= M_PROMISC;
743 	}
744 
745 	if ((m->m_flags & M_PROMISC) == 0) {
746 		if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
747 			return;
748 		if (m == NULL)
749 			return;
750 
751 		eh = mtod(m, struct ether_header *);
752 		etype = ntohs(eh->ether_type);
753 	}
754 
755 #if NAGR > 0
756 	if (ifp->if_agrprivate &&
757 	    __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) {
758 		m->m_flags &= ~M_PROMISC;
759 		agr_input(ifp, m);
760 		return;
761 	}
762 #endif
763 
764 	/* Handle input from a lagg(4) port */
765 	if (ifp->if_type == IFT_IEEE8023ADLAG) {
766 		KASSERT(lagg_input_ethernet_p != NULL);
767 		m = (*lagg_input_ethernet_p)(ifp, m);
768 		if (m == NULL)
769 			return;
770 	}
771 
772 	/*
773 	 * If VLANs are configured on the interface, check to
774 	 * see if the device performed the decapsulation and
775 	 * provided us with the tag.
776 	 */
777 	if (ec->ec_nvlans && vlan_has_tag(m)) {
778 #if NVLAN > 0
779 		/*
780 		 * vlan_input() will either recursively call ether_input()
781 		 * or drop the packet.
782 		 */
783 		vlan_input(ifp, m);
784 		return;
785 #else
786 		goto drop;
787 #endif
788 	}
789 
790 	/*
791 	 * Handle protocols that expect to have the Ethernet header
792 	 * (and possibly FCS) intact.
793 	 */
794 	switch (etype) {
795 	case ETHERTYPE_VLAN: {
796 		struct ether_vlan_header *evl = (void *)eh;
797 
798 		/*
799 		 * If there is a tag of 0, then the VLAN header was probably
800 		 * just being used to store the priority.  Extract the ether
801 		 * type, and if IP or IPV6, let them deal with it.
802 		 */
803 		if (m->m_len >= sizeof(*evl) &&
804 		    EVL_VLANOFTAG(evl->evl_tag) == 0) {
805 			etype = ntohs(evl->evl_proto);
806 			ehlen = sizeof(*evl);
807 			if ((m->m_flags & M_PROMISC) == 0 &&
808 			    (etype == ETHERTYPE_IP ||
809 			     etype == ETHERTYPE_IPV6))
810 				break;
811 		}
812 
813 #if NVLAN > 0
814 		/*
815 		 * vlan_input() will either recursively call ether_input()
816 		 * or drop the packet.
817 		 */
818 		if (ec->ec_nvlans != 0) {
819 			vlan_input(ifp, m);
820 			return;
821 		} else
822 #endif
823 			goto drop;
824 	}
825 
826 #if NPPPOE > 0
827 	case ETHERTYPE_PPPOEDISC:
828 		pppoedisc_input(ifp, m);
829 		return;
830 
831 	case ETHERTYPE_PPPOE:
832 		pppoe_input(ifp, m);
833 		return;
834 #endif
835 
836 	case ETHERTYPE_SLOWPROTOCOLS: {
837 		uint8_t subtype;
838 
839 		if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype))
840 			goto drop;
841 
842 		m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype);
843 		switch (subtype) {
844 #if NAGR > 0
845 		case SLOWPROTOCOLS_SUBTYPE_LACP:
846 			if (ifp->if_agrprivate) {
847 				ieee8023ad_lacp_input(ifp, m);
848 				return;
849 			}
850 			break;
851 
852 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
853 			if (ifp->if_agrprivate) {
854 				ieee8023ad_marker_input(ifp, m);
855 				return;
856 			}
857 			break;
858 #endif
859 
860 		default:
861 			if (subtype == 0 || subtype > 10) {
862 				/* illegal value */
863 				goto drop;
864 			}
865 			/* unknown subtype */
866 			break;
867 		}
868 	}
869 	/* FALLTHROUGH */
870 	default:
871 		if (m->m_flags & M_PROMISC)
872 			goto drop;
873 	}
874 
875 	/* If the CRC is still on the packet, trim it off. */
876 	if (m->m_flags & M_HASFCS) {
877 		m_adj(m, -ETHER_CRC_LEN);
878 		m->m_flags &= ~M_HASFCS;
879 	}
880 
881 	/* etype represents the size of the payload in this case */
882 	if (etype <= ETHERMTU + sizeof(struct ether_header)) {
883 		KASSERT(ehlen == sizeof(*eh));
884 #if defined (LLC) || defined (NETATALK)
885 		ether_input_llc(ifp, m, eh);
886 		return;
887 #else
888 		goto drop;
889 #endif
890 	}
891 
892 	/* Strip off the Ethernet header. */
893 	m_adj(m, ehlen);
894 
895 	switch (etype) {
896 #ifdef INET
897 	case ETHERTYPE_IP:
898 #ifdef GATEWAY
899 		if (ipflow_fastforward(m))
900 			return;
901 #endif
902 		pktq = ip_pktq;
903 		break;
904 
905 	case ETHERTYPE_ARP:
906 		isr = NETISR_ARP;
907 		inq = &arpintrq;
908 		break;
909 
910 	case ETHERTYPE_REVARP:
911 		revarpinput(m);	/* XXX queue? */
912 		return;
913 #endif
914 
915 #ifdef INET6
916 	case ETHERTYPE_IPV6:
917 		if (__predict_false(!in6_present))
918 			goto drop;
919 #ifdef GATEWAY
920 		if (ip6flow_fastforward(&m))
921 			return;
922 #endif
923 		pktq = ip6_pktq;
924 		break;
925 #endif
926 
927 #ifdef NETATALK
928 	case ETHERTYPE_ATALK:
929 		isr = NETISR_ATALK;
930 		inq = &atintrq1;
931 		break;
932 
933 	case ETHERTYPE_AARP:
934 		aarpinput(ifp, m); /* XXX queue? */
935 		return;
936 #endif
937 
938 #ifdef MPLS
939 	case ETHERTYPE_MPLS:
940 		isr = NETISR_MPLS;
941 		inq = &mplsintrq;
942 		break;
943 #endif
944 
945 	default:
946 		goto drop;
947 	}
948 
949 	if (__predict_true(pktq)) {
950 #ifdef NET_MPSAFE
951 		const u_int h = curcpu()->ci_index;
952 #else
953 		const uint32_t h = pktq_rps_hash(m);
954 #endif
955 		if (__predict_false(!pktq_enqueue(pktq, m, h))) {
956 			m_freem(m);
957 		}
958 		return;
959 	}
960 
961 	if (__predict_false(!inq)) {
962 		/* Should not happen. */
963 		goto drop;
964 	}
965 
966 	IFQ_ENQUEUE_ISR(inq, m, isr);
967 	return;
968 
969 drop:
970 	m_freem(m);
971 dropped:
972 	if_statinc(ifp, if_ierrors); /* XXX should have a dedicated counter? */
973 }
974 
975 /*
976  * Convert Ethernet address to printable (loggable) representation.
977  */
978 char *
979 ether_sprintf(const u_char *ap)
980 {
981 	static char etherbuf[3 * ETHER_ADDR_LEN];
982 	return ether_snprintf(etherbuf, sizeof(etherbuf), ap);
983 }
984 
985 char *
986 ether_snprintf(char *buf, size_t len, const u_char *ap)
987 {
988 	char *cp = buf;
989 	size_t i;
990 
991 	for (i = 0; i < len / 3; i++) {
992 		*cp++ = hexdigits[*ap >> 4];
993 		*cp++ = hexdigits[*ap++ & 0xf];
994 		*cp++ = ':';
995 	}
996 	*--cp = '\0';
997 	return buf;
998 }
999 
1000 static void
1001 ether_link_state_changed(struct ifnet *ifp, int link_state)
1002 {
1003 #if NVLAN > 0
1004 	struct ethercom *ec = (void *)ifp;
1005 
1006 	if (ec->ec_nvlans)
1007 		vlan_link_state_changed(ifp, link_state);
1008 #endif
1009 }
1010 
1011 /*
1012  * Perform common duties while attaching to interface list
1013  */
1014 void
1015 ether_ifattach(struct ifnet *ifp, const uint8_t *lla)
1016 {
1017 	struct ethercom *ec = (struct ethercom *)ifp;
1018 
1019 	ifp->if_type = IFT_ETHER;
1020 	ifp->if_hdrlen = ETHER_HDR_LEN;
1021 	ifp->if_dlt = DLT_EN10MB;
1022 	ifp->if_mtu = ETHERMTU;
1023 	ifp->if_output = ether_output;
1024 	ifp->_if_input = ether_input;
1025 	ifp->if_link_state_changed = ether_link_state_changed;
1026 	if (ifp->if_baudrate == 0)
1027 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
1028 
1029 	if (lla != NULL)
1030 		if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla));
1031 
1032 	LIST_INIT(&ec->ec_multiaddrs);
1033 	SIMPLEQ_INIT(&ec->ec_vids);
1034 	ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
1035 	ec->ec_flags = 0;
1036 	ifp->if_broadcastaddr = etherbroadcastaddr;
1037 	bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header));
1038 #ifdef MBUFTRACE
1039 	mowner_init_owner(&ec->ec_tx_mowner, ifp->if_xname, "tx");
1040 	mowner_init_owner(&ec->ec_rx_mowner, ifp->if_xname, "rx");
1041 	MOWNER_ATTACH(&ec->ec_tx_mowner);
1042 	MOWNER_ATTACH(&ec->ec_rx_mowner);
1043 	ifp->if_mowner = &ec->ec_tx_mowner;
1044 #endif
1045 }
1046 
1047 void
1048 ether_ifdetach(struct ifnet *ifp)
1049 {
1050 	struct ethercom *ec = (void *) ifp;
1051 	struct ether_multi *enm;
1052 
1053 	IFNET_ASSERT_UNLOCKED(ifp);
1054 	/*
1055 	 * Prevent further calls to ioctl (for example turning off
1056 	 * promiscuous mode from the bridge code), which eventually can
1057 	 * call if_init() which can cause panics because the interface
1058 	 * is in the process of being detached. Return device not configured
1059 	 * instead.
1060 	 */
1061 	ifp->if_ioctl = __FPTRCAST(int (*)(struct ifnet *, u_long, void *),
1062 	    enxio);
1063 
1064 #if NBRIDGE > 0
1065 	if (ifp->if_bridge)
1066 		bridge_ifdetach(ifp);
1067 #endif
1068 	bpf_detach(ifp);
1069 #if NVLAN > 0
1070 	if (ec->ec_nvlans)
1071 		vlan_ifdetach(ifp);
1072 #endif
1073 
1074 #if NLAGG > 0
1075 	if (ifp->if_lagg)
1076 		lagg_ifdetach(ifp);
1077 #endif
1078 
1079 	ETHER_LOCK(ec);
1080 	KASSERT(ec->ec_nvlans == 0);
1081 	while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) {
1082 		LIST_REMOVE(enm, enm_list);
1083 		kmem_free(enm, sizeof(*enm));
1084 		ec->ec_multicnt--;
1085 	}
1086 	ETHER_UNLOCK(ec);
1087 
1088 	mutex_obj_free(ec->ec_lock);
1089 	ec->ec_lock = NULL;
1090 
1091 	ifp->if_mowner = NULL;
1092 	MOWNER_DETACH(&ec->ec_rx_mowner);
1093 	MOWNER_DETACH(&ec->ec_tx_mowner);
1094 }
1095 
1096 #if 0
1097 /*
1098  * This is for reference.  We have a table-driven version
1099  * of the little-endian crc32 generator, which is faster
1100  * than the double-loop.
1101  */
1102 uint32_t
1103 ether_crc32_le(const uint8_t *buf, size_t len)
1104 {
1105 	uint32_t c, crc, carry;
1106 	size_t i, j;
1107 
1108 	crc = 0xffffffffU;	/* initial value */
1109 
1110 	for (i = 0; i < len; i++) {
1111 		c = buf[i];
1112 		for (j = 0; j < 8; j++) {
1113 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1114 			crc >>= 1;
1115 			c >>= 1;
1116 			if (carry)
1117 				crc = (crc ^ ETHER_CRC_POLY_LE);
1118 		}
1119 	}
1120 
1121 	return (crc);
1122 }
1123 #else
1124 uint32_t
1125 ether_crc32_le(const uint8_t *buf, size_t len)
1126 {
1127 	static const uint32_t crctab[] = {
1128 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1129 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1130 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1131 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1132 	};
1133 	uint32_t crc;
1134 	size_t i;
1135 
1136 	crc = 0xffffffffU;	/* initial value */
1137 
1138 	for (i = 0; i < len; i++) {
1139 		crc ^= buf[i];
1140 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1141 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1142 	}
1143 
1144 	return (crc);
1145 }
1146 #endif
1147 
1148 uint32_t
1149 ether_crc32_be(const uint8_t *buf, size_t len)
1150 {
1151 	uint32_t c, crc, carry;
1152 	size_t i, j;
1153 
1154 	crc = 0xffffffffU;	/* initial value */
1155 
1156 	for (i = 0; i < len; i++) {
1157 		c = buf[i];
1158 		for (j = 0; j < 8; j++) {
1159 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1160 			crc <<= 1;
1161 			c >>= 1;
1162 			if (carry)
1163 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1164 		}
1165 	}
1166 
1167 	return (crc);
1168 }
1169 
1170 #ifdef INET
1171 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] =
1172     { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 };
1173 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] =
1174     { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff };
1175 #endif
1176 #ifdef INET6
1177 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] =
1178     { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 };
1179 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] =
1180     { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff };
1181 #endif
1182 
1183 /*
1184  * ether_aton implementation, not using a static buffer.
1185  */
1186 int
1187 ether_aton_r(u_char *dest, size_t len, const char *str)
1188 {
1189 	const u_char *cp = (const void *)str;
1190 	u_char *ep;
1191 
1192 #define atox(c)	(((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10))
1193 
1194 	if (len < ETHER_ADDR_LEN)
1195 		return ENOSPC;
1196 
1197 	ep = dest + ETHER_ADDR_LEN;
1198 
1199 	while (*cp) {
1200 		if (!isxdigit(*cp))
1201 			return EINVAL;
1202 
1203 		*dest = atox(*cp);
1204 		cp++;
1205 		if (isxdigit(*cp)) {
1206 			*dest = (*dest << 4) | atox(*cp);
1207 			cp++;
1208 		}
1209 		dest++;
1210 
1211 		if (dest == ep)
1212 			return (*cp == '\0') ? 0 : ENAMETOOLONG;
1213 
1214 		switch (*cp) {
1215 		case ':':
1216 		case '-':
1217 		case '.':
1218 			cp++;
1219 			break;
1220 		}
1221 	}
1222 	return ENOBUFS;
1223 }
1224 
1225 /*
1226  * Convert a sockaddr into an Ethernet address or range of Ethernet
1227  * addresses.
1228  */
1229 int
1230 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN],
1231     uint8_t addrhi[ETHER_ADDR_LEN])
1232 {
1233 #ifdef INET
1234 	const struct sockaddr_in *sin;
1235 #endif
1236 #ifdef INET6
1237 	const struct sockaddr_in6 *sin6;
1238 #endif
1239 
1240 	switch (sa->sa_family) {
1241 
1242 	case AF_UNSPEC:
1243 		memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN);
1244 		memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1245 		break;
1246 
1247 #ifdef INET
1248 	case AF_INET:
1249 		sin = satocsin(sa);
1250 		if (sin->sin_addr.s_addr == INADDR_ANY) {
1251 			/*
1252 			 * An IP address of INADDR_ANY means listen to
1253 			 * or stop listening to all of the Ethernet
1254 			 * multicast addresses used for IP.
1255 			 * (This is for the sake of IP multicast routers.)
1256 			 */
1257 			memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN);
1258 			memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN);
1259 		} else {
1260 			ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo);
1261 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1262 		}
1263 		break;
1264 #endif
1265 #ifdef INET6
1266 	case AF_INET6:
1267 		sin6 = satocsin6(sa);
1268 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1269 			/*
1270 			 * An IP6 address of 0 means listen to or stop
1271 			 * listening to all of the Ethernet multicast
1272 			 * address used for IP6.
1273 			 * (This is used for multicast routers.)
1274 			 */
1275 			memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN);
1276 			memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN);
1277 		} else {
1278 			ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo);
1279 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1280 		}
1281 		break;
1282 #endif
1283 
1284 	default:
1285 		return EAFNOSUPPORT;
1286 	}
1287 	return 0;
1288 }
1289 
1290 /*
1291  * Add an Ethernet multicast address or range of addresses to the list for a
1292  * given interface.
1293  */
1294 int
1295 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec)
1296 {
1297 	struct ether_multi *enm, *_enm;
1298 	u_char addrlo[ETHER_ADDR_LEN];
1299 	u_char addrhi[ETHER_ADDR_LEN];
1300 	int error = 0;
1301 
1302 	/* Allocate out of lock */
1303 	enm = kmem_alloc(sizeof(*enm), KM_SLEEP);
1304 
1305 	ETHER_LOCK(ec);
1306 	error = ether_multiaddr(sa, addrlo, addrhi);
1307 	if (error != 0)
1308 		goto out;
1309 
1310 	/*
1311 	 * Verify that we have valid Ethernet multicast addresses.
1312 	 */
1313 	if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) {
1314 		error = EINVAL;
1315 		goto out;
1316 	}
1317 
1318 	/*
1319 	 * See if the address range is already in the list.
1320 	 */
1321 	_enm = ether_lookup_multi(addrlo, addrhi, ec);
1322 	if (_enm != NULL) {
1323 		/*
1324 		 * Found it; just increment the reference count.
1325 		 */
1326 		++_enm->enm_refcount;
1327 		error = 0;
1328 		goto out;
1329 	}
1330 
1331 	/*
1332 	 * Link a new multicast record into the interface's multicast list.
1333 	 */
1334 	memcpy(enm->enm_addrlo, addrlo, ETHER_ADDR_LEN);
1335 	memcpy(enm->enm_addrhi, addrhi, ETHER_ADDR_LEN);
1336 	enm->enm_refcount = 1;
1337 	LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list);
1338 	ec->ec_multicnt++;
1339 
1340 	/*
1341 	 * Return ENETRESET to inform the driver that the list has changed
1342 	 * and its reception filter should be adjusted accordingly.
1343 	 */
1344 	error = ENETRESET;
1345 	enm = NULL;
1346 
1347 out:
1348 	ETHER_UNLOCK(ec);
1349 	if (enm != NULL)
1350 		kmem_free(enm, sizeof(*enm));
1351 	return error;
1352 }
1353 
1354 /*
1355  * Delete a multicast address record.
1356  */
1357 int
1358 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec)
1359 {
1360 	struct ether_multi *enm;
1361 	u_char addrlo[ETHER_ADDR_LEN];
1362 	u_char addrhi[ETHER_ADDR_LEN];
1363 	int error;
1364 
1365 	ETHER_LOCK(ec);
1366 	error = ether_multiaddr(sa, addrlo, addrhi);
1367 	if (error != 0)
1368 		goto error;
1369 
1370 	/*
1371 	 * Look up the address in our list.
1372 	 */
1373 	enm = ether_lookup_multi(addrlo, addrhi, ec);
1374 	if (enm == NULL) {
1375 		error = ENXIO;
1376 		goto error;
1377 	}
1378 	if (--enm->enm_refcount != 0) {
1379 		/*
1380 		 * Still some claims to this record.
1381 		 */
1382 		error = 0;
1383 		goto error;
1384 	}
1385 
1386 	/*
1387 	 * No remaining claims to this record; unlink and free it.
1388 	 */
1389 	LIST_REMOVE(enm, enm_list);
1390 	ec->ec_multicnt--;
1391 	ETHER_UNLOCK(ec);
1392 	kmem_free(enm, sizeof(*enm));
1393 
1394 	/*
1395 	 * Return ENETRESET to inform the driver that the list has changed
1396 	 * and its reception filter should be adjusted accordingly.
1397 	 */
1398 	return ENETRESET;
1399 
1400 error:
1401 	ETHER_UNLOCK(ec);
1402 	return error;
1403 }
1404 
1405 void
1406 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb)
1407 {
1408 	ec->ec_ifflags_cb = cb;
1409 }
1410 
1411 void
1412 ether_set_vlan_cb(struct ethercom *ec, ether_vlancb_t cb)
1413 {
1414 
1415 	ec->ec_vlan_cb = cb;
1416 }
1417 
1418 static int
1419 ether_ioctl_reinit(struct ethercom *ec)
1420 {
1421 	struct ifnet *ifp = &ec->ec_if;
1422 	int error;
1423 
1424 	switch (ifp->if_flags & (IFF_UP | IFF_RUNNING)) {
1425 	case IFF_RUNNING:
1426 		/*
1427 		 * If interface is marked down and it is running,
1428 		 * then stop and disable it.
1429 		 */
1430 		(*ifp->if_stop)(ifp, 1);
1431 		break;
1432 	case IFF_UP:
1433 		/*
1434 		 * If interface is marked up and it is stopped, then
1435 		 * start it.
1436 		 */
1437 		return (*ifp->if_init)(ifp);
1438 	case IFF_UP | IFF_RUNNING:
1439 		error = 0;
1440 		if (ec->ec_ifflags_cb != NULL) {
1441 			error = (*ec->ec_ifflags_cb)(ec);
1442 			if (error == ENETRESET) {
1443 				/*
1444 				 * Reset the interface to pick up
1445 				 * changes in any other flags that
1446 				 * affect the hardware state.
1447 				 */
1448 				return (*ifp->if_init)(ifp);
1449 			}
1450 		} else
1451 			error = (*ifp->if_init)(ifp);
1452 		return error;
1453 	case 0:
1454 		break;
1455 	}
1456 
1457 	return 0;
1458 }
1459 
1460 /*
1461  * Common ioctls for Ethernet interfaces.  Note, we must be
1462  * called at splnet().
1463  */
1464 int
1465 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data)
1466 {
1467 	struct ethercom *ec = (void *)ifp;
1468 	struct eccapreq *eccr;
1469 	struct ifreq *ifr = (struct ifreq *)data;
1470 	struct if_laddrreq *iflr = data;
1471 	const struct sockaddr_dl *sdl;
1472 	static const uint8_t zero[ETHER_ADDR_LEN];
1473 	int error;
1474 
1475 	switch (cmd) {
1476 	case SIOCINITIFADDR:
1477 	    {
1478 		struct ifaddr *ifa = (struct ifaddr *)data;
1479 		if (ifa->ifa_addr->sa_family != AF_LINK
1480 		    && (ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
1481 		       (IFF_UP | IFF_RUNNING)) {
1482 			ifp->if_flags |= IFF_UP;
1483 			if ((error = (*ifp->if_init)(ifp)) != 0)
1484 				return error;
1485 		}
1486 #ifdef INET
1487 		if (ifa->ifa_addr->sa_family == AF_INET)
1488 			arp_ifinit(ifp, ifa);
1489 #endif
1490 		return 0;
1491 	    }
1492 
1493 	case SIOCSIFMTU:
1494 	    {
1495 		int maxmtu;
1496 
1497 		if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU)
1498 			maxmtu = ETHERMTU_JUMBO;
1499 		else
1500 			maxmtu = ETHERMTU;
1501 
1502 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu)
1503 			return EINVAL;
1504 		else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET)
1505 			return error;
1506 		else if (ifp->if_flags & IFF_UP) {
1507 			/* Make sure the device notices the MTU change. */
1508 			return (*ifp->if_init)(ifp);
1509 		} else
1510 			return 0;
1511 	    }
1512 
1513 	case SIOCSIFFLAGS:
1514 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
1515 			return error;
1516 		return ether_ioctl_reinit(ec);
1517 	case SIOCGIFFLAGS:
1518 		error = ifioctl_common(ifp, cmd, data);
1519 		if (error == 0) {
1520 			/* Set IFF_ALLMULTI for backcompat */
1521 			ifr->ifr_flags |= (ec->ec_flags & ETHER_F_ALLMULTI) ?
1522 			    IFF_ALLMULTI : 0;
1523 		}
1524 		return error;
1525 	case SIOCGETHERCAP:
1526 		eccr = (struct eccapreq *)data;
1527 		eccr->eccr_capabilities = ec->ec_capabilities;
1528 		eccr->eccr_capenable = ec->ec_capenable;
1529 		return 0;
1530 	case SIOCSETHERCAP:
1531 		eccr = (struct eccapreq *)data;
1532 		if ((eccr->eccr_capenable & ~ec->ec_capabilities) != 0)
1533 			return EINVAL;
1534 		if (eccr->eccr_capenable == ec->ec_capenable)
1535 			return 0;
1536 #if 0 /* notyet */
1537 		ec->ec_capenable = (ec->ec_capenable & ETHERCAP_CANTCHANGE)
1538 		    | (eccr->eccr_capenable & ~ETHERCAP_CANTCHANGE);
1539 #else
1540 		ec->ec_capenable = eccr->eccr_capenable;
1541 #endif
1542 		return ether_ioctl_reinit(ec);
1543 	case SIOCADDMULTI:
1544 		return ether_addmulti(ifreq_getaddr(cmd, ifr), ec);
1545 	case SIOCDELMULTI:
1546 		return ether_delmulti(ifreq_getaddr(cmd, ifr), ec);
1547 	case SIOCSIFMEDIA:
1548 	case SIOCGIFMEDIA:
1549 		if (ec->ec_mii != NULL)
1550 			return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media,
1551 			    cmd);
1552 		else if (ec->ec_ifmedia != NULL)
1553 			return ifmedia_ioctl(ifp, ifr, ec->ec_ifmedia, cmd);
1554 		else
1555 			return ENOTTY;
1556 		break;
1557 	case SIOCALIFADDR:
1558 		sdl = satocsdl(sstocsa(&iflr->addr));
1559 		if (sdl->sdl_family != AF_LINK)
1560 			;
1561 		else if (ETHER_IS_MULTICAST(CLLADDR(sdl)))
1562 			return EINVAL;
1563 		else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0)
1564 			return EINVAL;
1565 		/*FALLTHROUGH*/
1566 	default:
1567 		return ifioctl_common(ifp, cmd, data);
1568 	}
1569 	return 0;
1570 }
1571 
1572 /*
1573  * Enable/disable passing VLAN packets if the parent interface supports it.
1574  * Return:
1575  * 	 0: Ok
1576  *	-1: Parent interface does not support vlans
1577  *	>0: Error
1578  */
1579 int
1580 ether_enable_vlan_mtu(struct ifnet *ifp)
1581 {
1582 	int error;
1583 	struct ethercom *ec = (void *)ifp;
1584 
1585 	/* Parent does not support VLAN's */
1586 	if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0)
1587 		return -1;
1588 
1589 	/*
1590 	 * Parent supports the VLAN_MTU capability,
1591 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames;
1592 	 * enable it.
1593 	 */
1594 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1595 
1596 	/* Interface is down, defer for later */
1597 	if ((ifp->if_flags & IFF_UP) == 0)
1598 		return 0;
1599 
1600 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1601 		return 0;
1602 
1603 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1604 	return error;
1605 }
1606 
1607 int
1608 ether_disable_vlan_mtu(struct ifnet *ifp)
1609 {
1610 	int error;
1611 	struct ethercom *ec = (void *)ifp;
1612 
1613 	/* We still have VLAN's, defer for later */
1614 	if (ec->ec_nvlans != 0)
1615 		return 0;
1616 
1617 	/* Parent does not support VLAB's, nothing to do. */
1618 	if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0)
1619 		return -1;
1620 
1621 	/*
1622 	 * Disable Tx/Rx of VLAN-sized frames.
1623 	 */
1624 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1625 
1626 	/* Interface is down, defer for later */
1627 	if ((ifp->if_flags & IFF_UP) == 0)
1628 		return 0;
1629 
1630 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1631 		return 0;
1632 
1633 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1634 	return error;
1635 }
1636 
1637 static int
1638 ether_multicast_sysctl(SYSCTLFN_ARGS)
1639 {
1640 	struct ether_multi *enm;
1641 	struct ifnet *ifp;
1642 	struct ethercom *ec;
1643 	int error = 0;
1644 	size_t written;
1645 	struct psref psref;
1646 	int bound;
1647 	unsigned int multicnt;
1648 	struct ether_multi_sysctl *addrs;
1649 	int i;
1650 
1651 	if (namelen != 1)
1652 		return EINVAL;
1653 
1654 	bound = curlwp_bind();
1655 	ifp = if_get_byindex(name[0], &psref);
1656 	if (ifp == NULL) {
1657 		error = ENODEV;
1658 		goto out;
1659 	}
1660 	if (ifp->if_type != IFT_ETHER) {
1661 		if_put(ifp, &psref);
1662 		*oldlenp = 0;
1663 		goto out;
1664 	}
1665 	ec = (struct ethercom *)ifp;
1666 
1667 	if (oldp == NULL) {
1668 		if_put(ifp, &psref);
1669 		*oldlenp = ec->ec_multicnt * sizeof(*addrs);
1670 		goto out;
1671 	}
1672 
1673 	/*
1674 	 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which
1675 	 * is sleepable, while holding it. Copy data to a local buffer first
1676 	 * with the lock taken and then call sysctl_copyout without holding it.
1677 	 */
1678 retry:
1679 	multicnt = ec->ec_multicnt;
1680 
1681 	if (multicnt == 0) {
1682 		if_put(ifp, &psref);
1683 		*oldlenp = 0;
1684 		goto out;
1685 	}
1686 
1687 	addrs = kmem_zalloc(sizeof(*addrs) * multicnt, KM_SLEEP);
1688 
1689 	ETHER_LOCK(ec);
1690 	if (multicnt != ec->ec_multicnt) {
1691 		/* The number of multicast addresses has changed */
1692 		ETHER_UNLOCK(ec);
1693 		kmem_free(addrs, sizeof(*addrs) * multicnt);
1694 		goto retry;
1695 	}
1696 
1697 	i = 0;
1698 	LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) {
1699 		struct ether_multi_sysctl *addr = &addrs[i];
1700 		addr->enm_refcount = enm->enm_refcount;
1701 		memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN);
1702 		memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN);
1703 		i++;
1704 	}
1705 	ETHER_UNLOCK(ec);
1706 
1707 	error = 0;
1708 	written = 0;
1709 	for (i = 0; i < multicnt; i++) {
1710 		struct ether_multi_sysctl *addr = &addrs[i];
1711 
1712 		if (written + sizeof(*addr) > *oldlenp)
1713 			break;
1714 		error = sysctl_copyout(l, addr, oldp, sizeof(*addr));
1715 		if (error)
1716 			break;
1717 		written += sizeof(*addr);
1718 		oldp = (char *)oldp + sizeof(*addr);
1719 	}
1720 	kmem_free(addrs, sizeof(*addrs) * multicnt);
1721 
1722 	if_put(ifp, &psref);
1723 
1724 	*oldlenp = written;
1725 out:
1726 	curlwp_bindx(bound);
1727 	return error;
1728 }
1729 
1730 static void
1731 ether_sysctl_setup(struct sysctllog **clog)
1732 {
1733 	const struct sysctlnode *rnode = NULL;
1734 
1735 	sysctl_createv(clog, 0, NULL, &rnode,
1736 		       CTLFLAG_PERMANENT,
1737 		       CTLTYPE_NODE, "ether",
1738 		       SYSCTL_DESCR("Ethernet-specific information"),
1739 		       NULL, 0, NULL, 0,
1740 		       CTL_NET, CTL_CREATE, CTL_EOL);
1741 
1742 	sysctl_createv(clog, 0, &rnode, NULL,
1743 		       CTLFLAG_PERMANENT,
1744 		       CTLTYPE_NODE, "multicast",
1745 		       SYSCTL_DESCR("multicast addresses"),
1746 		       ether_multicast_sysctl, 0, NULL, 0,
1747 		       CTL_CREATE, CTL_EOL);
1748 }
1749 
1750 void
1751 etherinit(void)
1752 {
1753 
1754 #ifdef DIAGNOSTIC
1755 	mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET);
1756 #endif
1757 	ether_sysctl_setup(NULL);
1758 }
1759