xref: /netbsd-src/sys/net/if_ethersubr.c (revision fb5eed702691094bd687fbf1ded189c87457cd35)
1 /*	$NetBSD: if_ethersubr.c,v 1.303 2021/11/08 16:50:05 christos 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 1.303 2021/11/08 16:50:05 christos Exp $");
65 
66 #ifdef _KERNEL_OPT
67 #include "opt_inet.h"
68 #include "opt_atalk.h"
69 #include "opt_mbuftrace.h"
70 #include "opt_mpls.h"
71 #include "opt_gateway.h"
72 #include "opt_pppoe.h"
73 #include "opt_net_mpsafe.h"
74 #endif
75 
76 #include "vlan.h"
77 #include "pppoe.h"
78 #include "bridge.h"
79 #include "arp.h"
80 #include "agr.h"
81 
82 #include <sys/sysctl.h>
83 #include <sys/mbuf.h>
84 #include <sys/mutex.h>
85 #include <sys/ioctl.h>
86 #include <sys/errno.h>
87 #include <sys/device.h>
88 #include <sys/entropy.h>
89 #include <sys/rndsource.h>
90 #include <sys/cpu.h>
91 #include <sys/kmem.h>
92 #include <sys/hook.h>
93 
94 #include <net/if.h>
95 #include <net/netisr.h>
96 #include <net/route.h>
97 #include <net/if_llc.h>
98 #include <net/if_dl.h>
99 #include <net/if_types.h>
100 #include <net/pktqueue.h>
101 
102 #include <net/if_media.h>
103 #include <dev/mii/mii.h>
104 #include <dev/mii/miivar.h>
105 
106 #if NARP == 0
107 /*
108  * XXX there should really be a way to issue this warning from within config(8)
109  */
110 #error You have included NETATALK or a pseudo-device in your configuration that depends on the presence of ethernet interfaces, but have no such interfaces configured. Check if you really need pseudo-device bridge, pppoe, vlan or options NETATALK.
111 #endif
112 
113 #include <net/bpf.h>
114 
115 #include <net/if_ether.h>
116 #include <net/if_vlanvar.h>
117 
118 #if NPPPOE > 0
119 #include <net/if_pppoe.h>
120 #endif
121 
122 #if NAGR > 0
123 #include <net/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 static pktq_rps_hash_func_t ether_pktq_rps_hash_p;
187 
188 /* if_lagg(4) support */
189 struct mbuf *(*lagg_input_ethernet_p)(struct ifnet *, struct mbuf *);
190 
191 static int ether_output(struct ifnet *, struct mbuf *,
192     const struct sockaddr *, const struct rtentry *);
193 
194 /*
195  * Ethernet output routine.
196  * Encapsulate a packet of type family for the local net.
197  * Assumes that ifp is actually pointer to ethercom structure.
198  */
199 static int
200 ether_output(struct ifnet * const ifp0, struct mbuf * const m0,
201     const struct sockaddr * const dst, const struct rtentry *rt)
202 {
203 	uint8_t esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
204 	uint16_t etype = 0;
205 	int error = 0, hdrcmplt = 0;
206 	struct mbuf *m = m0;
207 	struct mbuf *mcopy = NULL;
208 	struct ether_header *eh;
209 	struct ifnet *ifp = ifp0;
210 #ifdef INET
211 	struct arphdr *ah;
212 #endif
213 #ifdef NETATALK
214 	struct at_ifaddr *aa;
215 #endif
216 
217 #ifdef MBUFTRACE
218 	m_claimm(m, ifp->if_mowner);
219 #endif
220 
221 #if NCARP > 0
222 	if (ifp->if_type == IFT_CARP) {
223 		struct ifaddr *ifa;
224 		int s = pserialize_read_enter();
225 
226 		/* loop back if this is going to the carp interface */
227 		if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP &&
228 		    (ifa = ifa_ifwithaddr(dst)) != NULL) {
229 			if (ifa->ifa_ifp == ifp0) {
230 				pserialize_read_exit(s);
231 				return looutput(ifp0, m, dst, rt);
232 			}
233 		}
234 		pserialize_read_exit(s);
235 
236 		ifp = ifp->if_carpdev;
237 		/* ac = (struct arpcom *)ifp; */
238 
239 		if ((ifp0->if_flags & (IFF_UP | IFF_RUNNING)) !=
240 		    (IFF_UP | IFF_RUNNING))
241 			senderr(ENETDOWN);
242 	}
243 #endif
244 
245 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
246 		senderr(ENETDOWN);
247 
248 	switch (dst->sa_family) {
249 
250 #ifdef INET
251 	case AF_INET:
252 		if (m->m_flags & M_BCAST) {
253 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
254 		} else if (m->m_flags & M_MCAST) {
255 			ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst);
256 		} else {
257 			error = arpresolve(ifp0, rt, m, dst, edst, sizeof(edst));
258 			if (error)
259 				return (error == EWOULDBLOCK) ? 0 : error;
260 		}
261 		/* If broadcasting on a simplex interface, loopback a copy */
262 		if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
263 			mcopy = m_copypacket(m, M_DONTWAIT);
264 		etype = htons(ETHERTYPE_IP);
265 		break;
266 
267 	case AF_ARP:
268 		ah = mtod(m, struct arphdr *);
269 		if (m->m_flags & M_BCAST) {
270 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
271 		} else {
272 			void *tha = ar_tha(ah);
273 
274 			if (tha == NULL) {
275 				/* fake with ARPHRD_IEEE1394 */
276 				m_freem(m);
277 				return 0;
278 			}
279 			memcpy(edst, tha, sizeof(edst));
280 		}
281 
282 		ah->ar_hrd = htons(ARPHRD_ETHER);
283 
284 		switch (ntohs(ah->ar_op)) {
285 		case ARPOP_REVREQUEST:
286 		case ARPOP_REVREPLY:
287 			etype = htons(ETHERTYPE_REVARP);
288 			break;
289 
290 		case ARPOP_REQUEST:
291 		case ARPOP_REPLY:
292 		default:
293 			etype = htons(ETHERTYPE_ARP);
294 		}
295 		break;
296 #endif
297 
298 #ifdef INET6
299 	case AF_INET6:
300 		if (m->m_flags & M_BCAST) {
301 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
302 		} else if (m->m_flags & M_MCAST) {
303 			ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr,
304 			    edst);
305 		} else {
306 			error = nd6_resolve(ifp0, rt, m, dst, edst,
307 			    sizeof(edst));
308 			if (error)
309 				return (error == EWOULDBLOCK) ? 0 : error;
310 		}
311 		etype = htons(ETHERTYPE_IPV6);
312 		break;
313 #endif
314 
315 #ifdef NETATALK
316 	case AF_APPLETALK: {
317 		struct ifaddr *ifa;
318 		int s;
319 
320 		KERNEL_LOCK(1, NULL);
321 
322 		if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) {
323 			KERNEL_UNLOCK_ONE(NULL);
324 			return 0;
325 		}
326 
327 		/*
328 		 * ifaddr is the first thing in at_ifaddr
329 		 */
330 		s = pserialize_read_enter();
331 		ifa = at_ifawithnet((const struct sockaddr_at *)dst, ifp);
332 		if (ifa == NULL) {
333 			pserialize_read_exit(s);
334 			KERNEL_UNLOCK_ONE(NULL);
335 			senderr(EADDRNOTAVAIL);
336 		}
337 		aa = (struct at_ifaddr *)ifa;
338 
339 		/*
340 		 * In the phase 2 case, we need to prepend an mbuf for the
341 		 * llc header.
342 		 */
343 		if (aa->aa_flags & AFA_PHASE2) {
344 			struct llc llc;
345 
346 			M_PREPEND(m, sizeof(struct llc), M_DONTWAIT);
347 			if (m == NULL) {
348 				pserialize_read_exit(s);
349 				KERNEL_UNLOCK_ONE(NULL);
350 				senderr(ENOBUFS);
351 			}
352 
353 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
354 			llc.llc_control = LLC_UI;
355 			memcpy(llc.llc_snap_org_code, at_org_code,
356 			    sizeof(llc.llc_snap_org_code));
357 			llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK);
358 			memcpy(mtod(m, void *), &llc, sizeof(struct llc));
359 		} else {
360 			etype = htons(ETHERTYPE_ATALK);
361 		}
362 		pserialize_read_exit(s);
363 		KERNEL_UNLOCK_ONE(NULL);
364 		break;
365 	}
366 #endif /* NETATALK */
367 
368 	case pseudo_AF_HDRCMPLT:
369 		hdrcmplt = 1;
370 		memcpy(esrc,
371 		    ((const struct ether_header *)dst->sa_data)->ether_shost,
372 		    sizeof(esrc));
373 		/* FALLTHROUGH */
374 
375 	case AF_UNSPEC:
376 		memcpy(edst,
377 		    ((const struct ether_header *)dst->sa_data)->ether_dhost,
378 		    sizeof(edst));
379 		/* AF_UNSPEC doesn't swap the byte order of the ether_type. */
380 		etype = ((const struct ether_header *)dst->sa_data)->ether_type;
381 		break;
382 
383 	default:
384 		printf("%s: can't handle af%d\n", ifp->if_xname,
385 		    dst->sa_family);
386 		senderr(EAFNOSUPPORT);
387 	}
388 
389 #ifdef MPLS
390 	{
391 		struct m_tag *mtag;
392 		mtag = m_tag_find(m, PACKET_TAG_MPLS);
393 		if (mtag != NULL) {
394 			/* Having the tag itself indicates it's MPLS */
395 			etype = htons(ETHERTYPE_MPLS);
396 			m_tag_delete(m, mtag);
397 		}
398 	}
399 #endif
400 
401 	if (mcopy)
402 		(void)looutput(ifp, mcopy, dst, rt);
403 
404 	KASSERT((m->m_flags & M_PKTHDR) != 0);
405 
406 	/*
407 	 * If no ether type is set, this must be a 802.2 formatted packet.
408 	 */
409 	if (etype == 0)
410 		etype = htons(m->m_pkthdr.len);
411 
412 	/*
413 	 * Add local net header. If no space in first mbuf, allocate another.
414 	 */
415 	M_PREPEND(m, sizeof(struct ether_header), M_DONTWAIT);
416 	if (m == NULL)
417 		senderr(ENOBUFS);
418 
419 	eh = mtod(m, struct ether_header *);
420 	/* Note: etype is already in network byte order. */
421 	memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
422 	memcpy(eh->ether_dhost, edst, sizeof(edst));
423 	if (hdrcmplt) {
424 		memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost));
425 	} else {
426 	 	memcpy(eh->ether_shost, CLLADDR(ifp->if_sadl),
427 		    sizeof(eh->ether_shost));
428 	}
429 
430 #if NCARP > 0
431 	if (ifp0 != ifp && ifp0->if_type == IFT_CARP) {
432 	 	memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl),
433 		    sizeof(eh->ether_shost));
434 	}
435 #endif
436 
437 	if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0)
438 		return error;
439 	if (m == NULL)
440 		return 0;
441 
442 #if NBRIDGE > 0
443 	/*
444 	 * Bridges require special output handling.
445 	 */
446 	if (ifp->if_bridge)
447 		return bridge_output(ifp, m, NULL, NULL);
448 #endif
449 
450 #if NCARP > 0
451 	if (ifp != ifp0)
452 		if_statadd(ifp0, if_obytes, m->m_pkthdr.len + ETHER_HDR_LEN);
453 #endif
454 
455 #ifdef ALTQ
456 	KERNEL_LOCK(1, NULL);
457 	/*
458 	 * If ALTQ is enabled on the parent interface, do
459 	 * classification; the queueing discipline might not
460 	 * require classification, but might require the
461 	 * address family/header pointer in the pktattr.
462 	 */
463 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
464 		altq_etherclassify(&ifp->if_snd, m);
465 	KERNEL_UNLOCK_ONE(NULL);
466 #endif
467 	return ifq_enqueue(ifp, m);
468 
469 bad:
470 	if_statinc(ifp, if_oerrors);
471 	if (m)
472 		m_freem(m);
473 	return error;
474 }
475 
476 #ifdef ALTQ
477 /*
478  * This routine is a slight hack to allow a packet to be classified
479  * if the Ethernet headers are present.  It will go away when ALTQ's
480  * classification engine understands link headers.
481  *
482  * XXX: We may need to do m_pullups here. First to ensure struct ether_header
483  * is indeed contiguous, then to read the LLC and so on.
484  */
485 void
486 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m)
487 {
488 	struct ether_header *eh;
489 	struct mbuf *mtop = m;
490 	uint16_t ether_type;
491 	int hlen, af, hdrsize;
492 	void *hdr;
493 
494 	KASSERT((mtop->m_flags & M_PKTHDR) != 0);
495 
496 	hlen = ETHER_HDR_LEN;
497 	eh = mtod(m, struct ether_header *);
498 
499 	ether_type = htons(eh->ether_type);
500 
501 	if (ether_type < ETHERMTU) {
502 		/* LLC/SNAP */
503 		struct llc *llc = (struct llc *)(eh + 1);
504 		hlen += 8;
505 
506 		if (m->m_len < hlen ||
507 		    llc->llc_dsap != LLC_SNAP_LSAP ||
508 		    llc->llc_ssap != LLC_SNAP_LSAP ||
509 		    llc->llc_control != LLC_UI) {
510 			/* Not SNAP. */
511 			goto bad;
512 		}
513 
514 		ether_type = htons(llc->llc_un.type_snap.ether_type);
515 	}
516 
517 	switch (ether_type) {
518 	case ETHERTYPE_IP:
519 		af = AF_INET;
520 		hdrsize = 20;		/* sizeof(struct ip) */
521 		break;
522 
523 	case ETHERTYPE_IPV6:
524 		af = AF_INET6;
525 		hdrsize = 40;		/* sizeof(struct ip6_hdr) */
526 		break;
527 
528 	default:
529 		af = AF_UNSPEC;
530 		hdrsize = 0;
531 		break;
532 	}
533 
534 	while (m->m_len <= hlen) {
535 		hlen -= m->m_len;
536 		m = m->m_next;
537 		if (m == NULL)
538 			goto bad;
539 	}
540 
541 	if (m->m_len < (hlen + hdrsize)) {
542 		/*
543 		 * protocol header not in a single mbuf.
544 		 * We can't cope with this situation right
545 		 * now (but it shouldn't ever happen, really, anyhow).
546 		 */
547 #ifdef DEBUG
548 		printf("altq_etherclassify: headers span multiple mbufs: "
549 		    "%d < %d\n", m->m_len, (hlen + hdrsize));
550 #endif
551 		goto bad;
552 	}
553 
554 	m->m_data += hlen;
555 	m->m_len -= hlen;
556 
557 	hdr = mtod(m, void *);
558 
559 	if (ALTQ_NEEDS_CLASSIFY(ifq)) {
560 		mtop->m_pkthdr.pattr_class =
561 		    (*ifq->altq_classify)(ifq->altq_clfier, m, af);
562 	}
563 	mtop->m_pkthdr.pattr_af = af;
564 	mtop->m_pkthdr.pattr_hdr = hdr;
565 
566 	m->m_data -= hlen;
567 	m->m_len += hlen;
568 
569 	return;
570 
571 bad:
572 	mtop->m_pkthdr.pattr_class = NULL;
573 	mtop->m_pkthdr.pattr_hdr = NULL;
574 	mtop->m_pkthdr.pattr_af = AF_UNSPEC;
575 }
576 #endif /* ALTQ */
577 
578 #if defined (LLC) || defined (NETATALK)
579 static void
580 ether_input_llc(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh)
581 {
582 	struct ifqueue *inq = NULL;
583 	int isr = 0;
584 	struct llc *l;
585 
586 	if (m->m_len < sizeof(*eh) + sizeof(struct llc))
587 		goto drop;
588 
589 	l = (struct llc *)(eh+1);
590 	switch (l->llc_dsap) {
591 #ifdef NETATALK
592 	case LLC_SNAP_LSAP:
593 		switch (l->llc_control) {
594 		case LLC_UI:
595 			if (l->llc_ssap != LLC_SNAP_LSAP)
596 				goto drop;
597 
598 			if (memcmp(&(l->llc_snap_org_code)[0],
599 			    at_org_code, sizeof(at_org_code)) == 0 &&
600 			    ntohs(l->llc_snap_ether_type) ==
601 			    ETHERTYPE_ATALK) {
602 				inq = &atintrq2;
603 				m_adj(m, sizeof(struct ether_header)
604 				    + sizeof(struct llc));
605 				isr = NETISR_ATALK;
606 				break;
607 			}
608 
609 			if (memcmp(&(l->llc_snap_org_code)[0],
610 			    aarp_org_code,
611 			    sizeof(aarp_org_code)) == 0 &&
612 			    ntohs(l->llc_snap_ether_type) ==
613 			    ETHERTYPE_AARP) {
614 				m_adj(m, sizeof(struct ether_header)
615 				    + sizeof(struct llc));
616 				aarpinput(ifp, m); /* XXX queue? */
617 				return;
618 			}
619 
620 		default:
621 			goto drop;
622 		}
623 		break;
624 #endif
625 	default:
626 		goto drop;
627 	}
628 
629 	KASSERT(inq != NULL);
630 	IFQ_ENQUEUE_ISR(inq, m, isr);
631 	return;
632 
633 drop:
634 	m_freem(m);
635 	if_statinc(ifp, if_ierrors); /* XXX should have a dedicated counter? */
636 	return;
637 }
638 #endif /* defined (LLC) || defined (NETATALK) */
639 
640 /*
641  * Process a received Ethernet packet;
642  * the packet is in the mbuf chain m with
643  * the ether header.
644  */
645 void
646 ether_input(struct ifnet *ifp, struct mbuf *m)
647 {
648 	struct ethercom *ec = (struct ethercom *) ifp;
649 	pktqueue_t *pktq = NULL;
650 	struct ifqueue *inq = NULL;
651 	uint16_t etype;
652 	struct ether_header *eh;
653 	size_t ehlen;
654 	static int earlypkts;
655 	int isr = 0;
656 #if NAGR > 0
657 	void *agrprivate;
658 #endif
659 
660 	KASSERT(!cpu_intr_p());
661 	KASSERT((m->m_flags & M_PKTHDR) != 0);
662 
663 	if ((ifp->if_flags & IFF_UP) == 0)
664 		goto drop;
665 
666 #ifdef MBUFTRACE
667 	m_claimm(m, &ec->ec_rx_mowner);
668 #endif
669 
670 	if (__predict_false(m->m_len < sizeof(*eh))) {
671 		if ((m = m_pullup(m, sizeof(*eh))) == NULL) {
672 			if_statinc(ifp, if_ierrors);
673 			return;
674 		}
675 	}
676 
677 	eh = mtod(m, struct ether_header *);
678 	etype = ntohs(eh->ether_type);
679 	ehlen = sizeof(*eh);
680 
681 	if (__predict_false(earlypkts < 100 ||
682 		entropy_epoch() == (unsigned)-1)) {
683 		rnd_add_data(NULL, eh, ehlen, 0);
684 		earlypkts++;
685 	}
686 
687 	/*
688 	 * Determine if the packet is within its size limits. For MPLS the
689 	 * header length is variable, so we skip the check.
690 	 */
691 	if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len >
692 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
693 #ifdef DIAGNOSTIC
694 		mutex_enter(&bigpktpps_lock);
695 		if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count,
696 		    bigpktppslim)) {
697 			printf("%s: discarding oversize frame (len=%d)\n",
698 			    ifp->if_xname, m->m_pkthdr.len);
699 		}
700 		mutex_exit(&bigpktpps_lock);
701 #endif
702 		goto drop;
703 	}
704 
705 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
706 		/*
707 		 * If this is not a simplex interface, drop the packet
708 		 * if it came from us.
709 		 */
710 		if ((ifp->if_flags & IFF_SIMPLEX) == 0 &&
711 		    memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost,
712 		    ETHER_ADDR_LEN) == 0) {
713 			goto drop;
714 		}
715 
716 		if (memcmp(etherbroadcastaddr,
717 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
718 			m->m_flags |= M_BCAST;
719 		else
720 			m->m_flags |= M_MCAST;
721 		if_statinc(ifp, if_imcasts);
722 	}
723 
724 	/* If the CRC is still on the packet, trim it off. */
725 	if (m->m_flags & M_HASFCS) {
726 		m_adj(m, -ETHER_CRC_LEN);
727 		m->m_flags &= ~M_HASFCS;
728 	}
729 
730 	if_statadd(ifp, if_ibytes, m->m_pkthdr.len);
731 
732 #if NCARP > 0
733 	if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) {
734 		/*
735 		 * Clear M_PROMISC, in case the packet comes from a
736 		 * vlan.
737 		 */
738 		m->m_flags &= ~M_PROMISC;
739 		if (carp_input(m, (uint8_t *)&eh->ether_shost,
740 		    (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0)
741 			return;
742 	}
743 #endif
744 
745 	if ((m->m_flags & (M_BCAST | M_MCAST | M_PROMISC)) == 0 &&
746 	    (ifp->if_flags & IFF_PROMISC) != 0 &&
747 	    memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost,
748 	     ETHER_ADDR_LEN) != 0) {
749 		m->m_flags |= M_PROMISC;
750 	}
751 
752 	if ((m->m_flags & M_PROMISC) == 0) {
753 		if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
754 			return;
755 		if (m == NULL)
756 			return;
757 
758 		eh = mtod(m, struct ether_header *);
759 		etype = ntohs(eh->ether_type);
760 	}
761 
762 #if NAGR > 0
763 	if (ifp->if_type != IFT_IEEE8023ADLAG) {
764 		agrprivate = ifp->if_lagg;
765 	} else {
766 		agrprivate = NULL;
767 	}
768 	if (agrprivate != NULL &&
769 	    __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) {
770 		m->m_flags &= ~M_PROMISC;
771 		agr_input(ifp, m);
772 		return;
773 	}
774 #endif
775 
776 	/* Handle input from a lagg(4) port */
777 	if (ifp->if_type == IFT_IEEE8023ADLAG) {
778 		KASSERT(lagg_input_ethernet_p != NULL);
779 		m = (*lagg_input_ethernet_p)(ifp, m);
780 		if (m == NULL)
781 			return;
782 	}
783 
784 	/*
785 	 * If VLANs are configured on the interface, check to
786 	 * see if the device performed the decapsulation and
787 	 * provided us with the tag.
788 	 */
789 	if (ec->ec_nvlans && vlan_has_tag(m)) {
790 #if NVLAN > 0
791 		/*
792 		 * vlan_input() will either recursively call ether_input()
793 		 * or drop the packet.
794 		 */
795 		vlan_input(ifp, m);
796 		return;
797 #else
798 		goto drop;
799 #endif
800 	}
801 
802 	/*
803 	 * Handle protocols that expect to have the Ethernet header
804 	 * (and possibly FCS) intact.
805 	 */
806 	switch (etype) {
807 	case ETHERTYPE_VLAN: {
808 		struct ether_vlan_header *evl = (void *)eh;
809 
810 		/*
811 		 * If there is a tag of 0, then the VLAN header was probably
812 		 * just being used to store the priority.  Extract the ether
813 		 * type, and if IP or IPV6, let them deal with it.
814 		 */
815 		if (m->m_len >= sizeof(*evl) &&
816 		    EVL_VLANOFTAG(ntohs(evl->evl_tag)) == 0) {
817 			etype = ntohs(evl->evl_proto);
818 			ehlen = sizeof(*evl);
819 			if ((m->m_flags & M_PROMISC) == 0 &&
820 			    (etype == ETHERTYPE_IP ||
821 			     etype == ETHERTYPE_IPV6))
822 				break;
823 		}
824 
825 #if NVLAN > 0
826 		/*
827 		 * vlan_input() will either recursively call ether_input()
828 		 * or drop the packet.
829 		 */
830 		if (ec->ec_nvlans != 0) {
831 			vlan_input(ifp, m);
832 			return;
833 		} else
834 #endif
835 			goto drop;
836 	}
837 
838 #if NPPPOE > 0
839 	case ETHERTYPE_PPPOEDISC:
840 		pppoedisc_input(ifp, m);
841 		return;
842 
843 	case ETHERTYPE_PPPOE:
844 		pppoe_input(ifp, m);
845 		return;
846 #endif
847 
848 	case ETHERTYPE_SLOWPROTOCOLS: {
849 		uint8_t subtype;
850 
851 		if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype))
852 			goto drop;
853 
854 		m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype);
855 		switch (subtype) {
856 #if NAGR > 0
857 		case SLOWPROTOCOLS_SUBTYPE_LACP:
858 			if (agrprivate != NULL) {
859 				ieee8023ad_lacp_input(ifp, m);
860 				return;
861 			}
862 			break;
863 
864 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
865 			if (agrprivate != NULL) {
866 				ieee8023ad_marker_input(ifp, m);
867 				return;
868 			}
869 			break;
870 #endif
871 
872 		default:
873 			if (subtype == 0 || subtype > 10) {
874 				/* illegal value */
875 				goto error;
876 			}
877 			/* unknown subtype */
878 			break;
879 		}
880 	}
881 	/* FALLTHROUGH */
882 	default:
883 		if (m->m_flags & M_PROMISC)
884 			goto drop;
885 	}
886 
887 	/* If the CRC is still on the packet, trim it off. */
888 	if (m->m_flags & M_HASFCS) {
889 		m_adj(m, -ETHER_CRC_LEN);
890 		m->m_flags &= ~M_HASFCS;
891 	}
892 
893 	/* etype represents the size of the payload in this case */
894 	if (etype <= ETHERMTU + sizeof(struct ether_header)) {
895 		KASSERT(ehlen == sizeof(*eh));
896 #if defined (LLC) || defined (NETATALK)
897 		ether_input_llc(ifp, m, eh);
898 		return;
899 #else
900 		goto error;
901 #endif
902 	}
903 
904 	/* Strip off the Ethernet header. */
905 	m_adj(m, ehlen);
906 
907 	switch (etype) {
908 #ifdef INET
909 	case ETHERTYPE_IP:
910 #ifdef GATEWAY
911 		if (ipflow_fastforward(m))
912 			return;
913 #endif
914 		pktq = ip_pktq;
915 		break;
916 
917 	case ETHERTYPE_ARP:
918 		isr = NETISR_ARP;
919 		inq = &arpintrq;
920 		break;
921 
922 	case ETHERTYPE_REVARP:
923 		revarpinput(m);	/* XXX queue? */
924 		return;
925 #endif
926 
927 #ifdef INET6
928 	case ETHERTYPE_IPV6:
929 		if (__predict_false(!in6_present))
930 			goto drop;
931 #ifdef GATEWAY
932 		if (ip6flow_fastforward(&m))
933 			return;
934 #endif
935 		pktq = ip6_pktq;
936 		break;
937 #endif
938 
939 #ifdef NETATALK
940 	case ETHERTYPE_ATALK:
941 		isr = NETISR_ATALK;
942 		inq = &atintrq1;
943 		break;
944 
945 	case ETHERTYPE_AARP:
946 		aarpinput(ifp, m); /* XXX queue? */
947 		return;
948 #endif
949 
950 #ifdef MPLS
951 	case ETHERTYPE_MPLS:
952 		isr = NETISR_MPLS;
953 		inq = &mplsintrq;
954 		break;
955 #endif
956 
957 	default:
958 		goto drop;
959 	}
960 
961 	if (__predict_true(pktq)) {
962 		const uint32_t h = pktq_rps_hash(&ether_pktq_rps_hash_p, m);
963 		if (__predict_false(!pktq_enqueue(pktq, m, h))) {
964 			m_freem(m);
965 		}
966 		return;
967 	}
968 
969 	if (__predict_false(!inq)) {
970 		/* Should not happen. */
971 		goto error;
972 	}
973 
974 	IFQ_ENQUEUE_ISR(inq, m, isr);
975 	return;
976 
977 drop:
978 	m_freem(m);
979 	if_statinc(ifp, if_iqdrops); /* XXX should have a dedicated counter? */
980 	return;
981 error:
982 	m_freem(m);
983 	if_statinc(ifp, if_ierrors); /* XXX should have a dedicated counter? */
984 	return;
985 }
986 
987 /*
988  * Convert Ethernet address to printable (loggable) representation.
989  */
990 char *
991 ether_sprintf(const u_char *ap)
992 {
993 	static char etherbuf[3 * ETHER_ADDR_LEN];
994 	return ether_snprintf(etherbuf, sizeof(etherbuf), ap);
995 }
996 
997 char *
998 ether_snprintf(char *buf, size_t len, const u_char *ap)
999 {
1000 	char *cp = buf;
1001 	size_t i;
1002 
1003 	for (i = 0; i < len / 3; i++) {
1004 		*cp++ = hexdigits[*ap >> 4];
1005 		*cp++ = hexdigits[*ap++ & 0xf];
1006 		*cp++ = ':';
1007 	}
1008 	*--cp = '\0';
1009 	return buf;
1010 }
1011 
1012 /*
1013  * Perform common duties while attaching to interface list
1014  */
1015 void
1016 ether_ifattach(struct ifnet *ifp, const uint8_t *lla)
1017 {
1018 	struct ethercom *ec = (struct ethercom *)ifp;
1019 	char xnamebuf[HOOKNAMSIZ];
1020 
1021 	ifp->if_type = IFT_ETHER;
1022 	ifp->if_hdrlen = ETHER_HDR_LEN;
1023 	ifp->if_dlt = DLT_EN10MB;
1024 	ifp->if_mtu = ETHERMTU;
1025 	ifp->if_output = ether_output;
1026 	ifp->_if_input = ether_input;
1027 	if (ifp->if_baudrate == 0)
1028 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
1029 
1030 	if (lla != NULL)
1031 		if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla));
1032 
1033 	LIST_INIT(&ec->ec_multiaddrs);
1034 	SIMPLEQ_INIT(&ec->ec_vids);
1035 	ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
1036 	ec->ec_flags = 0;
1037 	ifp->if_broadcastaddr = etherbroadcastaddr;
1038 	bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header));
1039 	snprintf(xnamebuf, sizeof(xnamebuf),
1040 	    "%s-ether_ifdetachhooks", ifp->if_xname);
1041 	ec->ec_ifdetach_hooks = simplehook_create(IPL_NET, xnamebuf);
1042 #ifdef MBUFTRACE
1043 	mowner_init_owner(&ec->ec_tx_mowner, ifp->if_xname, "tx");
1044 	mowner_init_owner(&ec->ec_rx_mowner, ifp->if_xname, "rx");
1045 	MOWNER_ATTACH(&ec->ec_tx_mowner);
1046 	MOWNER_ATTACH(&ec->ec_rx_mowner);
1047 	ifp->if_mowner = &ec->ec_tx_mowner;
1048 #endif
1049 }
1050 
1051 void
1052 ether_ifdetach(struct ifnet *ifp)
1053 {
1054 	struct ethercom *ec = (void *) ifp;
1055 	struct ether_multi *enm;
1056 
1057 	IFNET_ASSERT_UNLOCKED(ifp);
1058 	/*
1059 	 * Prevent further calls to ioctl (for example turning off
1060 	 * promiscuous mode from the bridge code), which eventually can
1061 	 * call if_init() which can cause panics because the interface
1062 	 * is in the process of being detached. Return device not configured
1063 	 * instead.
1064 	 */
1065 	ifp->if_ioctl = __FPTRCAST(int (*)(struct ifnet *, u_long, void *),
1066 	    enxio);
1067 
1068 	simplehook_dohooks(ec->ec_ifdetach_hooks);
1069 	KASSERT(!simplehook_has_hooks(ec->ec_ifdetach_hooks));
1070 	simplehook_destroy(ec->ec_ifdetach_hooks);
1071 
1072 	bpf_detach(ifp);
1073 
1074 	ETHER_LOCK(ec);
1075 	KASSERT(ec->ec_nvlans == 0);
1076 	while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) {
1077 		LIST_REMOVE(enm, enm_list);
1078 		kmem_free(enm, sizeof(*enm));
1079 		ec->ec_multicnt--;
1080 	}
1081 	ETHER_UNLOCK(ec);
1082 
1083 	mutex_obj_free(ec->ec_lock);
1084 	ec->ec_lock = NULL;
1085 
1086 	ifp->if_mowner = NULL;
1087 	MOWNER_DETACH(&ec->ec_rx_mowner);
1088 	MOWNER_DETACH(&ec->ec_tx_mowner);
1089 }
1090 
1091 void *
1092 ether_ifdetachhook_establish(struct ifnet *ifp,
1093     void (*fn)(void *), void *arg)
1094 {
1095 	struct ethercom *ec;
1096 	khook_t *hk;
1097 
1098 	if (ifp->if_type != IFT_ETHER)
1099 		return NULL;
1100 
1101 	ec = (struct ethercom *)ifp;
1102 	hk = simplehook_establish(ec->ec_ifdetach_hooks,
1103 	    fn, arg);
1104 
1105 	return (void *)hk;
1106 }
1107 
1108 void
1109 ether_ifdetachhook_disestablish(struct ifnet *ifp,
1110     void *vhook, kmutex_t *lock)
1111 {
1112 	struct ethercom *ec;
1113 
1114 	if (vhook == NULL)
1115 		return;
1116 
1117 	ec = (struct ethercom *)ifp;
1118 	simplehook_disestablish(ec->ec_ifdetach_hooks, vhook, lock);
1119 }
1120 
1121 #if 0
1122 /*
1123  * This is for reference.  We have a table-driven version
1124  * of the little-endian crc32 generator, which is faster
1125  * than the double-loop.
1126  */
1127 uint32_t
1128 ether_crc32_le(const uint8_t *buf, size_t len)
1129 {
1130 	uint32_t c, crc, carry;
1131 	size_t i, j;
1132 
1133 	crc = 0xffffffffU;	/* initial value */
1134 
1135 	for (i = 0; i < len; i++) {
1136 		c = buf[i];
1137 		for (j = 0; j < 8; j++) {
1138 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1139 			crc >>= 1;
1140 			c >>= 1;
1141 			if (carry)
1142 				crc = (crc ^ ETHER_CRC_POLY_LE);
1143 		}
1144 	}
1145 
1146 	return (crc);
1147 }
1148 #else
1149 uint32_t
1150 ether_crc32_le(const uint8_t *buf, size_t len)
1151 {
1152 	static const uint32_t crctab[] = {
1153 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1154 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1155 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1156 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1157 	};
1158 	uint32_t crc;
1159 	size_t i;
1160 
1161 	crc = 0xffffffffU;	/* initial value */
1162 
1163 	for (i = 0; i < len; i++) {
1164 		crc ^= buf[i];
1165 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1166 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1167 	}
1168 
1169 	return (crc);
1170 }
1171 #endif
1172 
1173 uint32_t
1174 ether_crc32_be(const uint8_t *buf, size_t len)
1175 {
1176 	uint32_t c, crc, carry;
1177 	size_t i, j;
1178 
1179 	crc = 0xffffffffU;	/* initial value */
1180 
1181 	for (i = 0; i < len; i++) {
1182 		c = buf[i];
1183 		for (j = 0; j < 8; j++) {
1184 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1185 			crc <<= 1;
1186 			c >>= 1;
1187 			if (carry)
1188 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1189 		}
1190 	}
1191 
1192 	return (crc);
1193 }
1194 
1195 #ifdef INET
1196 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] =
1197     { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 };
1198 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] =
1199     { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff };
1200 #endif
1201 #ifdef INET6
1202 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] =
1203     { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 };
1204 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] =
1205     { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff };
1206 #endif
1207 
1208 /*
1209  * ether_aton implementation, not using a static buffer.
1210  */
1211 int
1212 ether_aton_r(u_char *dest, size_t len, const char *str)
1213 {
1214 	const u_char *cp = (const void *)str;
1215 	u_char *ep;
1216 
1217 #define atox(c)	(((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10))
1218 
1219 	if (len < ETHER_ADDR_LEN)
1220 		return ENOSPC;
1221 
1222 	ep = dest + ETHER_ADDR_LEN;
1223 
1224 	while (*cp) {
1225 		if (!isxdigit(*cp))
1226 			return EINVAL;
1227 
1228 		*dest = atox(*cp);
1229 		cp++;
1230 		if (isxdigit(*cp)) {
1231 			*dest = (*dest << 4) | atox(*cp);
1232 			cp++;
1233 		}
1234 		dest++;
1235 
1236 		if (dest == ep)
1237 			return (*cp == '\0') ? 0 : ENAMETOOLONG;
1238 
1239 		switch (*cp) {
1240 		case ':':
1241 		case '-':
1242 		case '.':
1243 			cp++;
1244 			break;
1245 		}
1246 	}
1247 	return ENOBUFS;
1248 }
1249 
1250 /*
1251  * Convert a sockaddr into an Ethernet address or range of Ethernet
1252  * addresses.
1253  */
1254 int
1255 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN],
1256     uint8_t addrhi[ETHER_ADDR_LEN])
1257 {
1258 #ifdef INET
1259 	const struct sockaddr_in *sin;
1260 #endif
1261 #ifdef INET6
1262 	const struct sockaddr_in6 *sin6;
1263 #endif
1264 
1265 	switch (sa->sa_family) {
1266 
1267 	case AF_UNSPEC:
1268 		memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN);
1269 		memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1270 		break;
1271 
1272 #ifdef INET
1273 	case AF_INET:
1274 		sin = satocsin(sa);
1275 		if (sin->sin_addr.s_addr == INADDR_ANY) {
1276 			/*
1277 			 * An IP address of INADDR_ANY means listen to
1278 			 * or stop listening to all of the Ethernet
1279 			 * multicast addresses used for IP.
1280 			 * (This is for the sake of IP multicast routers.)
1281 			 */
1282 			memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN);
1283 			memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN);
1284 		} else {
1285 			ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo);
1286 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1287 		}
1288 		break;
1289 #endif
1290 #ifdef INET6
1291 	case AF_INET6:
1292 		sin6 = satocsin6(sa);
1293 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1294 			/*
1295 			 * An IP6 address of 0 means listen to or stop
1296 			 * listening to all of the Ethernet multicast
1297 			 * address used for IP6.
1298 			 * (This is used for multicast routers.)
1299 			 */
1300 			memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN);
1301 			memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN);
1302 		} else {
1303 			ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo);
1304 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1305 		}
1306 		break;
1307 #endif
1308 
1309 	default:
1310 		return EAFNOSUPPORT;
1311 	}
1312 	return 0;
1313 }
1314 
1315 /*
1316  * Add an Ethernet multicast address or range of addresses to the list for a
1317  * given interface.
1318  */
1319 int
1320 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec)
1321 {
1322 	struct ether_multi *enm, *_enm;
1323 	u_char addrlo[ETHER_ADDR_LEN];
1324 	u_char addrhi[ETHER_ADDR_LEN];
1325 	int error = 0;
1326 
1327 	/* Allocate out of lock */
1328 	enm = kmem_alloc(sizeof(*enm), KM_SLEEP);
1329 
1330 	ETHER_LOCK(ec);
1331 	error = ether_multiaddr(sa, addrlo, addrhi);
1332 	if (error != 0)
1333 		goto out;
1334 
1335 	/*
1336 	 * Verify that we have valid Ethernet multicast addresses.
1337 	 */
1338 	if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) {
1339 		error = EINVAL;
1340 		goto out;
1341 	}
1342 
1343 	/*
1344 	 * See if the address range is already in the list.
1345 	 */
1346 	_enm = ether_lookup_multi(addrlo, addrhi, ec);
1347 	if (_enm != NULL) {
1348 		/*
1349 		 * Found it; just increment the reference count.
1350 		 */
1351 		++_enm->enm_refcount;
1352 		error = 0;
1353 		goto out;
1354 	}
1355 
1356 	/*
1357 	 * Link a new multicast record into the interface's multicast list.
1358 	 */
1359 	memcpy(enm->enm_addrlo, addrlo, ETHER_ADDR_LEN);
1360 	memcpy(enm->enm_addrhi, addrhi, ETHER_ADDR_LEN);
1361 	enm->enm_refcount = 1;
1362 	LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list);
1363 	ec->ec_multicnt++;
1364 
1365 	/*
1366 	 * Return ENETRESET to inform the driver that the list has changed
1367 	 * and its reception filter should be adjusted accordingly.
1368 	 */
1369 	error = ENETRESET;
1370 	enm = NULL;
1371 
1372 out:
1373 	ETHER_UNLOCK(ec);
1374 	if (enm != NULL)
1375 		kmem_free(enm, sizeof(*enm));
1376 	return error;
1377 }
1378 
1379 /*
1380  * Delete a multicast address record.
1381  */
1382 int
1383 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec)
1384 {
1385 	struct ether_multi *enm;
1386 	u_char addrlo[ETHER_ADDR_LEN];
1387 	u_char addrhi[ETHER_ADDR_LEN];
1388 	int error;
1389 
1390 	ETHER_LOCK(ec);
1391 	error = ether_multiaddr(sa, addrlo, addrhi);
1392 	if (error != 0)
1393 		goto error;
1394 
1395 	/*
1396 	 * Look up the address in our list.
1397 	 */
1398 	enm = ether_lookup_multi(addrlo, addrhi, ec);
1399 	if (enm == NULL) {
1400 		error = ENXIO;
1401 		goto error;
1402 	}
1403 	if (--enm->enm_refcount != 0) {
1404 		/*
1405 		 * Still some claims to this record.
1406 		 */
1407 		error = 0;
1408 		goto error;
1409 	}
1410 
1411 	/*
1412 	 * No remaining claims to this record; unlink and free it.
1413 	 */
1414 	LIST_REMOVE(enm, enm_list);
1415 	ec->ec_multicnt--;
1416 	ETHER_UNLOCK(ec);
1417 	kmem_free(enm, sizeof(*enm));
1418 
1419 	/*
1420 	 * Return ENETRESET to inform the driver that the list has changed
1421 	 * and its reception filter should be adjusted accordingly.
1422 	 */
1423 	return ENETRESET;
1424 
1425 error:
1426 	ETHER_UNLOCK(ec);
1427 	return error;
1428 }
1429 
1430 void
1431 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb)
1432 {
1433 	ec->ec_ifflags_cb = cb;
1434 }
1435 
1436 void
1437 ether_set_vlan_cb(struct ethercom *ec, ether_vlancb_t cb)
1438 {
1439 
1440 	ec->ec_vlan_cb = cb;
1441 }
1442 
1443 static int
1444 ether_ioctl_reinit(struct ethercom *ec)
1445 {
1446 	struct ifnet *ifp = &ec->ec_if;
1447 	int error;
1448 
1449 	switch (ifp->if_flags & (IFF_UP | IFF_RUNNING)) {
1450 	case IFF_RUNNING:
1451 		/*
1452 		 * If interface is marked down and it is running,
1453 		 * then stop and disable it.
1454 		 */
1455 		(*ifp->if_stop)(ifp, 1);
1456 		break;
1457 	case IFF_UP:
1458 		/*
1459 		 * If interface is marked up and it is stopped, then
1460 		 * start it.
1461 		 */
1462 		return (*ifp->if_init)(ifp);
1463 	case IFF_UP | IFF_RUNNING:
1464 		error = 0;
1465 		if (ec->ec_ifflags_cb != NULL) {
1466 			error = (*ec->ec_ifflags_cb)(ec);
1467 			if (error == ENETRESET) {
1468 				/*
1469 				 * Reset the interface to pick up
1470 				 * changes in any other flags that
1471 				 * affect the hardware state.
1472 				 */
1473 				return (*ifp->if_init)(ifp);
1474 			}
1475 		} else
1476 			error = (*ifp->if_init)(ifp);
1477 		return error;
1478 	case 0:
1479 		break;
1480 	}
1481 
1482 	return 0;
1483 }
1484 
1485 /*
1486  * Common ioctls for Ethernet interfaces.  Note, we must be
1487  * called at splnet().
1488  */
1489 int
1490 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data)
1491 {
1492 	struct ethercom *ec = (void *)ifp;
1493 	struct eccapreq *eccr;
1494 	struct ifreq *ifr = (struct ifreq *)data;
1495 	struct if_laddrreq *iflr = data;
1496 	const struct sockaddr_dl *sdl;
1497 	static const uint8_t zero[ETHER_ADDR_LEN];
1498 	int error;
1499 
1500 	switch (cmd) {
1501 	case SIOCINITIFADDR:
1502 	    {
1503 		struct ifaddr *ifa = (struct ifaddr *)data;
1504 		if (ifa->ifa_addr->sa_family != AF_LINK
1505 		    && (ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
1506 		       (IFF_UP | IFF_RUNNING)) {
1507 			ifp->if_flags |= IFF_UP;
1508 			if ((error = (*ifp->if_init)(ifp)) != 0)
1509 				return error;
1510 		}
1511 #ifdef INET
1512 		if (ifa->ifa_addr->sa_family == AF_INET)
1513 			arp_ifinit(ifp, ifa);
1514 #endif
1515 		return 0;
1516 	    }
1517 
1518 	case SIOCSIFMTU:
1519 	    {
1520 		int maxmtu;
1521 
1522 		if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU)
1523 			maxmtu = ETHERMTU_JUMBO;
1524 		else
1525 			maxmtu = ETHERMTU;
1526 
1527 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu)
1528 			return EINVAL;
1529 		else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET)
1530 			return error;
1531 		else if (ifp->if_flags & IFF_UP) {
1532 			/* Make sure the device notices the MTU change. */
1533 			return (*ifp->if_init)(ifp);
1534 		} else
1535 			return 0;
1536 	    }
1537 
1538 	case SIOCSIFFLAGS:
1539 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
1540 			return error;
1541 		return ether_ioctl_reinit(ec);
1542 	case SIOCGIFFLAGS:
1543 		error = ifioctl_common(ifp, cmd, data);
1544 		if (error == 0) {
1545 			/* Set IFF_ALLMULTI for backcompat */
1546 			ifr->ifr_flags |= (ec->ec_flags & ETHER_F_ALLMULTI) ?
1547 			    IFF_ALLMULTI : 0;
1548 		}
1549 		return error;
1550 	case SIOCGETHERCAP:
1551 		eccr = (struct eccapreq *)data;
1552 		eccr->eccr_capabilities = ec->ec_capabilities;
1553 		eccr->eccr_capenable = ec->ec_capenable;
1554 		return 0;
1555 	case SIOCSETHERCAP:
1556 		eccr = (struct eccapreq *)data;
1557 		if ((eccr->eccr_capenable & ~ec->ec_capabilities) != 0)
1558 			return EINVAL;
1559 		if (eccr->eccr_capenable == ec->ec_capenable)
1560 			return 0;
1561 #if 0 /* notyet */
1562 		ec->ec_capenable = (ec->ec_capenable & ETHERCAP_CANTCHANGE)
1563 		    | (eccr->eccr_capenable & ~ETHERCAP_CANTCHANGE);
1564 #else
1565 		ec->ec_capenable = eccr->eccr_capenable;
1566 #endif
1567 		return ether_ioctl_reinit(ec);
1568 	case SIOCADDMULTI:
1569 		return ether_addmulti(ifreq_getaddr(cmd, ifr), ec);
1570 	case SIOCDELMULTI:
1571 		return ether_delmulti(ifreq_getaddr(cmd, ifr), ec);
1572 	case SIOCSIFMEDIA:
1573 	case SIOCGIFMEDIA:
1574 		if (ec->ec_mii != NULL)
1575 			return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media,
1576 			    cmd);
1577 		else if (ec->ec_ifmedia != NULL)
1578 			return ifmedia_ioctl(ifp, ifr, ec->ec_ifmedia, cmd);
1579 		else
1580 			return ENOTTY;
1581 		break;
1582 	case SIOCALIFADDR:
1583 		sdl = satocsdl(sstocsa(&iflr->addr));
1584 		if (sdl->sdl_family != AF_LINK)
1585 			;
1586 		else if (ETHER_IS_MULTICAST(CLLADDR(sdl)))
1587 			return EINVAL;
1588 		else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0)
1589 			return EINVAL;
1590 		/*FALLTHROUGH*/
1591 	default:
1592 		return ifioctl_common(ifp, cmd, data);
1593 	}
1594 	return 0;
1595 }
1596 
1597 /*
1598  * Enable/disable passing VLAN packets if the parent interface supports it.
1599  * Return:
1600  * 	 0: Ok
1601  *	-1: Parent interface does not support vlans
1602  *	>0: Error
1603  */
1604 int
1605 ether_enable_vlan_mtu(struct ifnet *ifp)
1606 {
1607 	int error;
1608 	struct ethercom *ec = (void *)ifp;
1609 
1610 	/* Parent does not support VLAN's */
1611 	if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0)
1612 		return -1;
1613 
1614 	/*
1615 	 * Parent supports the VLAN_MTU capability,
1616 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames;
1617 	 * enable it.
1618 	 */
1619 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1620 
1621 	/* Interface is down, defer for later */
1622 	if ((ifp->if_flags & IFF_UP) == 0)
1623 		return 0;
1624 
1625 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1626 		return 0;
1627 
1628 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1629 	return error;
1630 }
1631 
1632 int
1633 ether_disable_vlan_mtu(struct ifnet *ifp)
1634 {
1635 	int error;
1636 	struct ethercom *ec = (void *)ifp;
1637 
1638 	/* We still have VLAN's, defer for later */
1639 	if (ec->ec_nvlans != 0)
1640 		return 0;
1641 
1642 	/* Parent does not support VLAB's, nothing to do. */
1643 	if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0)
1644 		return -1;
1645 
1646 	/*
1647 	 * Disable Tx/Rx of VLAN-sized frames.
1648 	 */
1649 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1650 
1651 	/* Interface is down, defer for later */
1652 	if ((ifp->if_flags & IFF_UP) == 0)
1653 		return 0;
1654 
1655 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1656 		return 0;
1657 
1658 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1659 	return error;
1660 }
1661 
1662 static int
1663 ether_multicast_sysctl(SYSCTLFN_ARGS)
1664 {
1665 	struct ether_multi *enm;
1666 	struct ifnet *ifp;
1667 	struct ethercom *ec;
1668 	int error = 0;
1669 	size_t written;
1670 	struct psref psref;
1671 	int bound;
1672 	unsigned int multicnt;
1673 	struct ether_multi_sysctl *addrs;
1674 	int i;
1675 
1676 	if (namelen != 1)
1677 		return EINVAL;
1678 
1679 	bound = curlwp_bind();
1680 	ifp = if_get_byindex(name[0], &psref);
1681 	if (ifp == NULL) {
1682 		error = ENODEV;
1683 		goto out;
1684 	}
1685 	if (ifp->if_type != IFT_ETHER) {
1686 		if_put(ifp, &psref);
1687 		*oldlenp = 0;
1688 		goto out;
1689 	}
1690 	ec = (struct ethercom *)ifp;
1691 
1692 	if (oldp == NULL) {
1693 		if_put(ifp, &psref);
1694 		*oldlenp = ec->ec_multicnt * sizeof(*addrs);
1695 		goto out;
1696 	}
1697 
1698 	/*
1699 	 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which
1700 	 * is sleepable, while holding it. Copy data to a local buffer first
1701 	 * with the lock taken and then call sysctl_copyout without holding it.
1702 	 */
1703 retry:
1704 	multicnt = ec->ec_multicnt;
1705 
1706 	if (multicnt == 0) {
1707 		if_put(ifp, &psref);
1708 		*oldlenp = 0;
1709 		goto out;
1710 	}
1711 
1712 	addrs = kmem_zalloc(sizeof(*addrs) * multicnt, KM_SLEEP);
1713 
1714 	ETHER_LOCK(ec);
1715 	if (multicnt != ec->ec_multicnt) {
1716 		/* The number of multicast addresses has changed */
1717 		ETHER_UNLOCK(ec);
1718 		kmem_free(addrs, sizeof(*addrs) * multicnt);
1719 		goto retry;
1720 	}
1721 
1722 	i = 0;
1723 	LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) {
1724 		struct ether_multi_sysctl *addr = &addrs[i];
1725 		addr->enm_refcount = enm->enm_refcount;
1726 		memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN);
1727 		memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN);
1728 		i++;
1729 	}
1730 	ETHER_UNLOCK(ec);
1731 
1732 	error = 0;
1733 	written = 0;
1734 	for (i = 0; i < multicnt; i++) {
1735 		struct ether_multi_sysctl *addr = &addrs[i];
1736 
1737 		if (written + sizeof(*addr) > *oldlenp)
1738 			break;
1739 		error = sysctl_copyout(l, addr, oldp, sizeof(*addr));
1740 		if (error)
1741 			break;
1742 		written += sizeof(*addr);
1743 		oldp = (char *)oldp + sizeof(*addr);
1744 	}
1745 	kmem_free(addrs, sizeof(*addrs) * multicnt);
1746 
1747 	if_put(ifp, &psref);
1748 
1749 	*oldlenp = written;
1750 out:
1751 	curlwp_bindx(bound);
1752 	return error;
1753 }
1754 
1755 static void
1756 ether_sysctl_setup(struct sysctllog **clog)
1757 {
1758 	const struct sysctlnode *rnode = NULL;
1759 
1760 	sysctl_createv(clog, 0, NULL, &rnode,
1761 		       CTLFLAG_PERMANENT,
1762 		       CTLTYPE_NODE, "ether",
1763 		       SYSCTL_DESCR("Ethernet-specific information"),
1764 		       NULL, 0, NULL, 0,
1765 		       CTL_NET, CTL_CREATE, CTL_EOL);
1766 
1767 	sysctl_createv(clog, 0, &rnode, NULL,
1768 		       CTLFLAG_PERMANENT,
1769 		       CTLTYPE_NODE, "multicast",
1770 		       SYSCTL_DESCR("multicast addresses"),
1771 		       ether_multicast_sysctl, 0, NULL, 0,
1772 		       CTL_CREATE, CTL_EOL);
1773 
1774 	sysctl_createv(clog, 0, &rnode, NULL,
1775 		       CTLFLAG_PERMANENT | CTLFLAG_READWRITE,
1776 		       CTLTYPE_STRING, "rps_hash",
1777 		       SYSCTL_DESCR("Interface rps hash function control"),
1778 		       sysctl_pktq_rps_hash_handler, 0, (void *)&ether_pktq_rps_hash_p,
1779 		       PKTQ_RPS_HASH_NAME_LEN,
1780 		       CTL_CREATE, CTL_EOL);
1781 }
1782 
1783 void
1784 etherinit(void)
1785 {
1786 
1787 #ifdef DIAGNOSTIC
1788 	mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET);
1789 #endif
1790 	ether_pktq_rps_hash_p = pktq_rps_hash_default;
1791 	ether_sysctl_setup(NULL);
1792 }
1793