xref: /netbsd-src/sys/net/if_ethersubr.c (revision 7c192b2a5e1093666e67801684f930ef49b3b363)
1 /*	$NetBSD: if_ethersubr.c,v 1.243 2017/07/23 10:55:00 para 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 1.243 2017/07/23 10:55:00 para Exp $");
65 
66 #ifdef _KERNEL_OPT
67 #include "opt_inet.h"
68 #include "opt_atalk.h"
69 #include "opt_mbuftrace.h"
70 #include "opt_mpls.h"
71 #include "opt_gateway.h"
72 #include "opt_pppoe.h"
73 #include "opt_net_mpsafe.h"
74 #endif
75 
76 #include "vlan.h"
77 #include "pppoe.h"
78 #include "bridge.h"
79 #include "arp.h"
80 #include "agr.h"
81 
82 #include <sys/sysctl.h>
83 #include <sys/mbuf.h>
84 #include <sys/mutex.h>
85 #include <sys/ioctl.h>
86 #include <sys/errno.h>
87 #include <sys/device.h>
88 #include <sys/rnd.h>
89 #include <sys/rndsource.h>
90 #include <sys/cpu.h>
91 #include <sys/kmem.h>
92 
93 #include <net/if.h>
94 #include <net/netisr.h>
95 #include <net/route.h>
96 #include <net/if_llc.h>
97 #include <net/if_dl.h>
98 #include <net/if_types.h>
99 #include <net/pktqueue.h>
100 
101 #include <net/if_media.h>
102 #include <dev/mii/mii.h>
103 #include <dev/mii/miivar.h>
104 
105 #if NARP == 0
106 /*
107  * XXX there should really be a way to issue this warning from within config(8)
108  */
109 #error You have included NETATALK or a pseudo-device in your configuration that depends on the presence of ethernet interfaces, but have no such interfaces configured. Check if you really need pseudo-device bridge, pppoe, vlan or options NETATALK.
110 #endif
111 
112 #include <net/bpf.h>
113 
114 #include <net/if_ether.h>
115 #include <net/if_vlanvar.h>
116 
117 #if NPPPOE > 0
118 #include <net/if_pppoe.h>
119 #endif
120 
121 #if NAGR > 0
122 #include <net/agr/ieee8023_slowprotocols.h>	/* XXX */
123 #include <net/agr/ieee8023ad.h>
124 #include <net/agr/if_agrvar.h>
125 #endif
126 
127 #if NBRIDGE > 0
128 #include <net/if_bridgevar.h>
129 #endif
130 
131 #include <netinet/in.h>
132 #ifdef INET
133 #include <netinet/in_var.h>
134 #endif
135 #include <netinet/if_inarp.h>
136 
137 #ifdef INET6
138 #ifndef INET
139 #include <netinet/in.h>
140 #endif
141 #include <netinet6/in6_var.h>
142 #include <netinet6/nd6.h>
143 #endif
144 
145 
146 #include "carp.h"
147 #if NCARP > 0
148 #include <netinet/ip_carp.h>
149 #endif
150 
151 #ifdef NETATALK
152 #include <netatalk/at.h>
153 #include <netatalk/at_var.h>
154 #include <netatalk/at_extern.h>
155 
156 #define llc_snap_org_code llc_un.type_snap.org_code
157 #define llc_snap_ether_type llc_un.type_snap.ether_type
158 
159 extern u_char	at_org_code[3];
160 extern u_char	aarp_org_code[3];
161 #endif /* NETATALK */
162 
163 #ifdef MPLS
164 #include <netmpls/mpls.h>
165 #include <netmpls/mpls_var.h>
166 #endif
167 
168 static struct timeval bigpktppslim_last;
169 static int bigpktppslim = 2;	/* XXX */
170 static int bigpktpps_count;
171 static kmutex_t bigpktpps_lock __cacheline_aligned;
172 
173 
174 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] =
175     { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
176 const uint8_t ethermulticastaddr_slowprotocols[ETHER_ADDR_LEN] =
177     { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x02 };
178 #define senderr(e) { error = (e); goto bad;}
179 
180 static	int ether_output(struct ifnet *, struct mbuf *,
181 	    const struct sockaddr *, const struct rtentry *);
182 
183 /*
184  * Ethernet output routine.
185  * Encapsulate a packet of type family for the local net.
186  * Assumes that ifp is actually pointer to ethercom structure.
187  */
188 static int
189 ether_output(struct ifnet * const ifp0, struct mbuf * const m0,
190 	const struct sockaddr * const dst,
191 	const struct rtentry *rt)
192 {
193 	uint16_t etype = 0;
194 	int error = 0, hdrcmplt = 0;
195  	uint8_t esrc[6], edst[6];
196 	struct mbuf *m = m0;
197 	struct mbuf *mcopy = NULL;
198 	struct ether_header *eh;
199 	struct ifnet *ifp = ifp0;
200 #ifdef INET
201 	struct arphdr *ah;
202 #endif /* INET */
203 #ifdef NETATALK
204 	struct at_ifaddr *aa;
205 #endif /* NETATALK */
206 
207 	/*
208 	 * some paths such as carp_output() call ethr_output() with "ifp"
209 	 * argument as other than ether ifnet.
210 	 */
211 	KASSERT(ifp->if_output != ether_output
212 	    || ifp->if_extflags & IFEF_OUTPUT_MPSAFE);
213 
214 #ifdef MBUFTRACE
215 	m_claimm(m, ifp->if_mowner);
216 #endif
217 
218 #if NCARP > 0
219 	if (ifp->if_type == IFT_CARP) {
220 		struct ifaddr *ifa;
221 		int s = pserialize_read_enter();
222 
223 		/* loop back if this is going to the carp interface */
224 		if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP &&
225 		    (ifa = ifa_ifwithaddr(dst)) != NULL) {
226 			if (ifa->ifa_ifp == ifp0) {
227 				pserialize_read_exit(s);
228 				return looutput(ifp0, m, dst, rt);
229 			}
230 		}
231 		pserialize_read_exit(s);
232 
233 		ifp = ifp->if_carpdev;
234 		/* ac = (struct arpcom *)ifp; */
235 
236 		if ((ifp0->if_flags & (IFF_UP|IFF_RUNNING)) !=
237 		    (IFF_UP|IFF_RUNNING))
238 			senderr(ENETDOWN);
239 	}
240 #endif /* NCARP > 0 */
241 
242 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
243 		senderr(ENETDOWN);
244 
245 	switch (dst->sa_family) {
246 
247 #ifdef INET
248 	case AF_INET:
249 		if (m->m_flags & M_BCAST)
250 			(void)memcpy(edst, etherbroadcastaddr, sizeof(edst));
251 		else if (m->m_flags & M_MCAST)
252 			ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst);
253 		else if ((error = arpresolve(ifp, rt, m, dst, edst,
254 		    sizeof(edst))) != 0) {
255 			return error == EWOULDBLOCK ? 0 : error;
256 		}
257 		/* If broadcasting on a simplex interface, loopback a copy */
258 		if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
259 			mcopy = m_copy(m, 0, (int)M_COPYALL);
260 		etype = htons(ETHERTYPE_IP);
261 		break;
262 
263 	case AF_ARP:
264 		ah = mtod(m, struct arphdr *);
265 		if (m->m_flags & M_BCAST)
266 			(void)memcpy(edst, etherbroadcastaddr, sizeof(edst));
267 		else {
268 			void *tha = ar_tha(ah);
269 
270 			if (tha == NULL) {
271 				/* fake with ARPHDR_IEEE1394 */
272 				m_freem(m);
273 				return 0;
274 			}
275 			memcpy(edst, tha, sizeof(edst));
276 		}
277 
278 		ah->ar_hrd = htons(ARPHRD_ETHER);
279 
280 		switch (ntohs(ah->ar_op)) {
281 		case ARPOP_REVREQUEST:
282 		case ARPOP_REVREPLY:
283 			etype = htons(ETHERTYPE_REVARP);
284 			break;
285 
286 		case ARPOP_REQUEST:
287 		case ARPOP_REPLY:
288 		default:
289 			etype = htons(ETHERTYPE_ARP);
290 		}
291 
292 		break;
293 #endif
294 #ifdef INET6
295 	case AF_INET6:
296 		if (m->m_flags & M_BCAST)
297 			(void)memcpy(edst, etherbroadcastaddr, sizeof(edst));
298 		else if (m->m_flags & M_MCAST) {
299 			ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr,
300 			    edst);
301 		} else {
302 			error = nd6_resolve(ifp, rt, m, dst, edst,
303 			    sizeof(edst));
304 			if (error != 0)
305 				return error == EWOULDBLOCK ? 0 : error;
306 		}
307 		etype = htons(ETHERTYPE_IPV6);
308 		break;
309 #endif
310 #ifdef NETATALK
311     case AF_APPLETALK: {
312 		struct ifaddr *ifa;
313 		int s;
314 
315 		KERNEL_LOCK(1, NULL);
316 		if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) {
317 #ifdef NETATALKDEBUG
318 			printf("aarpresolv failed\n");
319 #endif /* NETATALKDEBUG */
320 			KERNEL_UNLOCK_ONE(NULL);
321 			return (0);
322 		}
323 		/*
324 		 * ifaddr is the first thing in at_ifaddr
325 		 */
326 		s = pserialize_read_enter();
327 		ifa = at_ifawithnet((const struct sockaddr_at *)dst, ifp);
328 		if (ifa == NULL) {
329 			pserialize_read_exit(s);
330 			KERNEL_UNLOCK_ONE(NULL);
331 			goto bad;
332 		}
333 		aa = (struct at_ifaddr *)ifa;
334 
335 		/*
336 		 * In the phase 2 case, we need to prepend an mbuf for the
337 		 * llc header.  Since we must preserve the value of m,
338 		 * which is passed to us by value, we m_copy() the first
339 		 * mbuf, and use it for our 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 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
346 			llc.llc_control = LLC_UI;
347 			memcpy(llc.llc_snap_org_code, at_org_code,
348 			    sizeof(llc.llc_snap_org_code));
349 			llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK);
350 			memcpy(mtod(m, void *), &llc, sizeof(struct llc));
351 		} else {
352 			etype = htons(ETHERTYPE_ATALK);
353 		}
354 		pserialize_read_exit(s);
355 		KERNEL_UNLOCK_ONE(NULL);
356 		break;
357 	    }
358 #endif /* NETATALK */
359 	case pseudo_AF_HDRCMPLT:
360 		hdrcmplt = 1;
361 		memcpy(esrc,
362 		    ((const struct ether_header *)dst->sa_data)->ether_shost,
363 		    sizeof(esrc));
364 		/* FALLTHROUGH */
365 
366 	case AF_UNSPEC:
367  		memcpy(edst,
368 		    ((const struct ether_header *)dst->sa_data)->ether_dhost,
369 		    sizeof(edst));
370 		/* AF_UNSPEC doesn't swap the byte order of the ether_type. */
371 		etype = ((const struct ether_header *)dst->sa_data)->ether_type;
372 		break;
373 
374 	default:
375 		printf("%s: can't handle af%d\n", ifp->if_xname,
376 			dst->sa_family);
377 		senderr(EAFNOSUPPORT);
378 	}
379 
380 #ifdef MPLS
381 	KERNEL_LOCK(1, NULL);
382 	{
383 		struct m_tag *mtag;
384 		mtag = m_tag_find(m, PACKET_TAG_MPLS, NULL);
385 		if (mtag != NULL) {
386 			/* Having the tag itself indicates it's MPLS */
387 			etype = htons(ETHERTYPE_MPLS);
388 			m_tag_delete(m, mtag);
389 		}
390 	}
391 	KERNEL_UNLOCK_ONE(NULL);
392 #endif
393 
394 	if (mcopy)
395 		(void)looutput(ifp, mcopy, dst, rt);
396 
397 	/* If no ether type is set, this must be a 802.2 formatted packet.
398 	 */
399 	if (etype == 0)
400 		etype = htons(m->m_pkthdr.len);
401 	/*
402 	 * Add local net header.  If no space in first mbuf,
403 	 * allocate another.
404 	 */
405 	M_PREPEND(m, sizeof (struct ether_header), M_DONTWAIT);
406 	if (m == 0)
407 		senderr(ENOBUFS);
408 	eh = mtod(m, struct ether_header *);
409 	/* Note: etype is already in network byte order. */
410 	(void)memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
411  	memcpy(eh->ether_dhost, edst, sizeof(edst));
412 	if (hdrcmplt)
413 		memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost));
414 	else
415 	 	memcpy(eh->ether_shost, CLLADDR(ifp->if_sadl),
416 		    sizeof(eh->ether_shost));
417 
418 #if NCARP > 0
419 	if (ifp0 != ifp && ifp0->if_type == IFT_CARP) {
420 	 	memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl),
421 		    sizeof(eh->ether_shost));
422 	}
423 #endif /* NCARP > 0 */
424 
425 	if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0)
426 		return (error);
427 	if (m == NULL)
428 		return (0);
429 
430 #if NBRIDGE > 0
431 	/*
432 	 * Bridges require special output handling.
433 	 */
434 	if (ifp->if_bridge)
435 		return (bridge_output(ifp, m, NULL, NULL));
436 #endif
437 
438 #if NCARP > 0
439 	if (ifp != ifp0)
440 		ifp0->if_obytes += m->m_pkthdr.len + ETHER_HDR_LEN;
441 #endif /* NCARP > 0 */
442 
443 #ifdef ALTQ
444 	KERNEL_LOCK(1, NULL);
445 	/*
446 	 * If ALTQ is enabled on the parent interface, do
447 	 * classification; the queueing discipline might not
448 	 * require classification, but might require the
449 	 * address family/header pointer in the pktattr.
450 	 */
451 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
452 		altq_etherclassify(&ifp->if_snd, m);
453 	KERNEL_UNLOCK_ONE(NULL);
454 #endif
455 	return ifq_enqueue(ifp, m);
456 
457 bad:
458 	if (m)
459 		m_freem(m);
460 	return (error);
461 }
462 
463 #ifdef ALTQ
464 /*
465  * This routine is a slight hack to allow a packet to be classified
466  * if the Ethernet headers are present.  It will go away when ALTQ's
467  * classification engine understands link headers.
468  */
469 void
470 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m)
471 {
472 	struct ether_header *eh;
473 	uint16_t ether_type;
474 	int hlen, af, hdrsize;
475 	void *hdr;
476 
477 	hlen = ETHER_HDR_LEN;
478 	eh = mtod(m, struct ether_header *);
479 
480 	ether_type = htons(eh->ether_type);
481 
482 	if (ether_type < ETHERMTU) {
483 		/* LLC/SNAP */
484 		struct llc *llc = (struct llc *)(eh + 1);
485 		hlen += 8;
486 
487 		if (m->m_len < hlen ||
488 		    llc->llc_dsap != LLC_SNAP_LSAP ||
489 		    llc->llc_ssap != LLC_SNAP_LSAP ||
490 		    llc->llc_control != LLC_UI) {
491 			/* Not SNAP. */
492 			goto bad;
493 		}
494 
495 		ether_type = htons(llc->llc_un.type_snap.ether_type);
496 	}
497 
498 	switch (ether_type) {
499 	case ETHERTYPE_IP:
500 		af = AF_INET;
501 		hdrsize = 20;		/* sizeof(struct ip) */
502 		break;
503 
504 	case ETHERTYPE_IPV6:
505 		af = AF_INET6;
506 		hdrsize = 40;		/* sizeof(struct ip6_hdr) */
507 		break;
508 
509 	default:
510 		af = AF_UNSPEC;
511 		hdrsize = 0;
512 		break;
513 	}
514 
515 	while (m->m_len <= hlen) {
516 		hlen -= m->m_len;
517 		m = m->m_next;
518 	}
519 	if (m->m_len < (hlen + hdrsize)) {
520 		/*
521 		 * protocol header not in a single mbuf.
522 		 * We can't cope with this situation right
523 		 * now (but it shouldn't ever happen, really, anyhow).
524 		 */
525 #ifdef DEBUG
526 		printf("altq_etherclassify: headers span multiple mbufs: "
527 		    "%d < %d\n", m->m_len, (hlen + hdrsize));
528 #endif
529 		goto bad;
530 	}
531 
532 	m->m_data += hlen;
533 	m->m_len -= hlen;
534 
535 	hdr = mtod(m, void *);
536 
537 	if (ALTQ_NEEDS_CLASSIFY(ifq))
538 		m->m_pkthdr.pattr_class =
539 		    (*ifq->altq_classify)(ifq->altq_clfier, m, af);
540 	m->m_pkthdr.pattr_af = af;
541 	m->m_pkthdr.pattr_hdr = hdr;
542 
543 	m->m_data -= hlen;
544 	m->m_len += hlen;
545 
546 	return;
547 
548  bad:
549 	m->m_pkthdr.pattr_class = NULL;
550 	m->m_pkthdr.pattr_hdr = NULL;
551 	m->m_pkthdr.pattr_af = AF_UNSPEC;
552 }
553 #endif /* ALTQ */
554 
555 /*
556  * Process a received Ethernet packet;
557  * the packet is in the mbuf chain m with
558  * the ether header.
559  */
560 void
561 ether_input(struct ifnet *ifp, struct mbuf *m)
562 {
563 	struct ethercom *ec = (struct ethercom *) ifp;
564 	pktqueue_t *pktq = NULL;
565 	struct ifqueue *inq = NULL;
566 	uint16_t etype;
567 	struct ether_header *eh;
568 	size_t ehlen;
569 	static int earlypkts;
570 	int isr = 0;
571 #if defined (LLC) || defined(NETATALK)
572 	struct llc *l;
573 #endif
574 
575 	KASSERT(!cpu_intr_p());
576 
577 	if ((ifp->if_flags & IFF_UP) == 0) {
578 		m_freem(m);
579 		return;
580 	}
581 
582 #ifdef MBUFTRACE
583 	m_claimm(m, &ec->ec_rx_mowner);
584 #endif
585 	eh = mtod(m, struct ether_header *);
586 	etype = ntohs(eh->ether_type);
587 	ehlen = sizeof(*eh);
588 
589 	if(__predict_false(earlypkts < 100 || !rnd_initial_entropy)) {
590 		rnd_add_data(NULL, eh, ehlen, 0);
591 		earlypkts++;
592 	}
593 
594 	/*
595 	 * Determine if the packet is within its size limits.
596 	 */
597 	if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len >
598 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
599 		mutex_enter(&bigpktpps_lock);
600 		if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count,
601 			    bigpktppslim)) {
602 			printf("%s: discarding oversize frame (len=%d)\n",
603 			    ifp->if_xname, m->m_pkthdr.len);
604 		}
605 		mutex_exit(&bigpktpps_lock);
606 		m_freem(m);
607 		return;
608 	}
609 
610 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
611 		/*
612 		 * If this is not a simplex interface, drop the packet
613 		 * if it came from us.
614 		 */
615 		if ((ifp->if_flags & IFF_SIMPLEX) == 0 &&
616 		    memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost,
617 		    ETHER_ADDR_LEN) == 0) {
618 			m_freem(m);
619 			return;
620 		}
621 
622 		if (memcmp(etherbroadcastaddr,
623 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
624 			m->m_flags |= M_BCAST;
625 		else
626 			m->m_flags |= M_MCAST;
627 		ifp->if_imcasts++;
628 	}
629 
630 	/* If the CRC is still on the packet, trim it off. */
631 	if (m->m_flags & M_HASFCS) {
632 		m_adj(m, -ETHER_CRC_LEN);
633 		m->m_flags &= ~M_HASFCS;
634 	}
635 
636 	ifp->if_ibytes += m->m_pkthdr.len;
637 
638 #if NCARP > 0
639 	if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) {
640 		/*
641 		 * clear M_PROMISC, in case the packets comes from a
642 		 * vlan
643 		 */
644 		m->m_flags &= ~M_PROMISC;
645 		if (carp_input(m, (uint8_t *)&eh->ether_shost,
646 		    (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0)
647 			return;
648 	}
649 #endif /* NCARP > 0 */
650 	if ((m->m_flags & (M_BCAST|M_MCAST|M_PROMISC)) == 0 &&
651 	    (ifp->if_flags & IFF_PROMISC) != 0 &&
652 	    memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost,
653 		   ETHER_ADDR_LEN) != 0) {
654 		m->m_flags |= M_PROMISC;
655 	}
656 
657 	if ((m->m_flags & M_PROMISC) == 0) {
658 		if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
659 			return;
660 		if (m == NULL)
661 			return;
662 
663 		eh = mtod(m, struct ether_header *);
664 		etype = ntohs(eh->ether_type);
665 		ehlen = sizeof(*eh);
666 	}
667 
668 #if NAGR > 0
669 	if (ifp->if_agrprivate &&
670 	    __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) {
671 		m->m_flags &= ~M_PROMISC;
672 		agr_input(ifp, m);
673 		return;
674 	}
675 #endif /* NAGR > 0 */
676 
677 	/*
678 	 * If VLANs are configured on the interface, check to
679 	 * see if the device performed the decapsulation and
680 	 * provided us with the tag.
681 	 */
682 	if (ec->ec_nvlans && m_tag_find(m, PACKET_TAG_VLAN, NULL) != NULL) {
683 #if NVLAN > 0
684 		/*
685 		 * vlan_input() will either recursively call ether_input()
686 		 * or drop the packet.
687 		 */
688 		vlan_input(ifp, m);
689 #else
690 		m_freem(m);
691 #endif
692 		return;
693 	}
694 
695 	/*
696 	 * Handle protocols that expect to have the Ethernet header
697 	 * (and possibly FCS) intact.
698 	 */
699 	switch (etype) {
700 	case ETHERTYPE_VLAN: {
701 		struct ether_vlan_header *evl = (void *)eh;
702 		/*
703 		 * If there is a tag of 0, then the VLAN header was probably
704 		 * just being used to store the priority.  Extract the ether
705 		 * type, and if IP or IPV6, let them deal with it.
706 		 */
707 		if (m->m_len <= sizeof(*evl)
708 		    && EVL_VLANOFTAG(evl->evl_tag) == 0) {
709 			etype = ntohs(evl->evl_proto);
710 			ehlen = sizeof(*evl);
711 			if ((m->m_flags & M_PROMISC) == 0
712 			    && (etype == ETHERTYPE_IP
713 				|| etype == ETHERTYPE_IPV6))
714 				break;
715 		}
716 #if NVLAN > 0
717 		/*
718 		 * vlan_input() will either recursively call ether_input()
719 		 * or drop the packet.
720 		 */
721 		if (((struct ethercom *)ifp)->ec_nvlans != 0)
722 			vlan_input(ifp, m);
723 		else
724 #endif /* NVLAN > 0 */
725 			m_freem(m);
726 		return;
727 	}
728 #if NPPPOE > 0
729 	case ETHERTYPE_PPPOEDISC:
730 		pppoedisc_input(ifp, m);
731 		return;
732 	case ETHERTYPE_PPPOE:
733 		pppoe_input(ifp, m);
734 		return;
735 #endif /* NPPPOE > 0 */
736 	case ETHERTYPE_SLOWPROTOCOLS: {
737 		uint8_t subtype;
738 
739 #if defined(DIAGNOSTIC)
740 		if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype)) {
741 			panic("ether_input: too short slow protocol packet");
742 		}
743 #endif
744 		m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype);
745 		switch (subtype) {
746 #if NAGR > 0
747 		case SLOWPROTOCOLS_SUBTYPE_LACP:
748 			if (ifp->if_agrprivate) {
749 				ieee8023ad_lacp_input(ifp, m);
750 				return;
751 			}
752 			break;
753 
754 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
755 			if (ifp->if_agrprivate) {
756 				ieee8023ad_marker_input(ifp, m);
757 				return;
758 			}
759 			break;
760 #endif /* NAGR > 0 */
761 		default:
762 			if (subtype == 0 || subtype > 10) {
763 				/* illegal value */
764 				m_freem(m);
765 				return;
766 			}
767 			/* unknown subtype */
768 			break;
769 		}
770 		/* FALLTHROUGH */
771 	}
772 	default:
773 		if (m->m_flags & M_PROMISC) {
774 			m_freem(m);
775 			return;
776 		}
777 	}
778 
779 	/* If the CRC is still on the packet, trim it off. */
780 	if (m->m_flags & M_HASFCS) {
781 		m_adj(m, -ETHER_CRC_LEN);
782 		m->m_flags &= ~M_HASFCS;
783 	}
784 
785 	if (etype > ETHERMTU + sizeof (struct ether_header)) {
786 		/* Strip off the Ethernet header. */
787 		m_adj(m, ehlen);
788 
789 		switch (etype) {
790 #ifdef INET
791 		case ETHERTYPE_IP:
792 #ifdef GATEWAY
793 			if (ipflow_fastforward(m))
794 				return;
795 #endif
796 			pktq = ip_pktq;
797 			break;
798 
799 		case ETHERTYPE_ARP:
800 			isr = NETISR_ARP;
801 			inq = &arpintrq;
802 			break;
803 
804 		case ETHERTYPE_REVARP:
805 			revarpinput(m);	/* XXX queue? */
806 			return;
807 #endif
808 #ifdef INET6
809 		case ETHERTYPE_IPV6:
810 			if (__predict_false(!in6_present)) {
811 				m_freem(m);
812 				return;
813 			}
814 #ifdef GATEWAY
815 			if (ip6flow_fastforward(&m))
816 				return;
817 #endif
818 			pktq = ip6_pktq;
819 			break;
820 #endif
821 #ifdef NETATALK
822 		case ETHERTYPE_ATALK:
823 			isr = NETISR_ATALK;
824 			inq = &atintrq1;
825 			break;
826 		case ETHERTYPE_AARP:
827 			/* probably this should be done with a NETISR as well */
828 			aarpinput(ifp, m); /* XXX */
829 			return;
830 #endif /* NETATALK */
831 #ifdef MPLS
832 		case ETHERTYPE_MPLS:
833 			isr = NETISR_MPLS;
834 			inq = &mplsintrq;
835 			break;
836 #endif
837 		default:
838 			m_freem(m);
839 			return;
840 		}
841 	} else {
842 #if defined (LLC) || defined (NETATALK)
843 		l = (struct llc *)(eh+1);
844 		switch (l->llc_dsap) {
845 #ifdef NETATALK
846 		case LLC_SNAP_LSAP:
847 			switch (l->llc_control) {
848 			case LLC_UI:
849 				if (l->llc_ssap != LLC_SNAP_LSAP) {
850 					goto dropanyway;
851 				}
852 
853 				if (memcmp(&(l->llc_snap_org_code)[0],
854 				    at_org_code, sizeof(at_org_code)) == 0 &&
855 				    ntohs(l->llc_snap_ether_type) ==
856 				    ETHERTYPE_ATALK) {
857 					inq = &atintrq2;
858 					m_adj(m, sizeof(struct ether_header)
859 					    + sizeof(struct llc));
860 					isr = NETISR_ATALK;
861 					break;
862 				}
863 
864 				if (memcmp(&(l->llc_snap_org_code)[0],
865 				    aarp_org_code,
866 				    sizeof(aarp_org_code)) == 0 &&
867 				    ntohs(l->llc_snap_ether_type) ==
868 				    ETHERTYPE_AARP) {
869 					m_adj( m, sizeof(struct ether_header)
870 					    + sizeof(struct llc));
871 					aarpinput(ifp, m); /* XXX */
872 				    return;
873 				}
874 
875 			default:
876 				goto dropanyway;
877 			}
878 			break;
879 		dropanyway:
880 #endif
881 		default:
882 			m_freem(m);
883 			return;
884 		}
885 #else /* ISO || LLC || NETATALK*/
886 		m_freem(m);
887 		return;
888 #endif /* ISO || LLC || NETATALK*/
889 	}
890 
891 	if (__predict_true(pktq)) {
892 #ifdef NET_MPSAFE
893 		const u_int h = curcpu()->ci_index;
894 #else
895 		const uint32_t h = pktq_rps_hash(m);
896 #endif
897 		if (__predict_false(!pktq_enqueue(pktq, m, h))) {
898 			m_freem(m);
899 		}
900 		return;
901 	}
902 
903 	if (__predict_false(!inq)) {
904 		/* Should not happen. */
905 		m_freem(m);
906 		return;
907 	}
908 
909 	IFQ_LOCK(inq);
910 	if (IF_QFULL(inq)) {
911 		IF_DROP(inq);
912 		IFQ_UNLOCK(inq);
913 		m_freem(m);
914 	} else {
915 		IF_ENQUEUE(inq, m);
916 		IFQ_UNLOCK(inq);
917 		schednetisr(isr);
918 	}
919 }
920 
921 /*
922  * Convert Ethernet address to printable (loggable) representation.
923  */
924 char *
925 ether_sprintf(const u_char *ap)
926 {
927 	static char etherbuf[3 * ETHER_ADDR_LEN];
928 	return ether_snprintf(etherbuf, sizeof(etherbuf), ap);
929 }
930 
931 char *
932 ether_snprintf(char *buf, size_t len, const u_char *ap)
933 {
934 	char *cp = buf;
935 	size_t i;
936 
937 	for (i = 0; i < len / 3; i++) {
938 		*cp++ = hexdigits[*ap >> 4];
939 		*cp++ = hexdigits[*ap++ & 0xf];
940 		*cp++ = ':';
941 	}
942 	*--cp = '\0';
943 	return buf;
944 }
945 
946 /*
947  * Perform common duties while attaching to interface list
948  */
949 void
950 ether_ifattach(struct ifnet *ifp, const uint8_t *lla)
951 {
952 	struct ethercom *ec = (struct ethercom *)ifp;
953 
954 	ifp->if_extflags |= IFEF_OUTPUT_MPSAFE;
955 	ifp->if_type = IFT_ETHER;
956 	ifp->if_hdrlen = ETHER_HDR_LEN;
957 	ifp->if_dlt = DLT_EN10MB;
958 	ifp->if_mtu = ETHERMTU;
959 	ifp->if_output = ether_output;
960 	ifp->_if_input = ether_input;
961 	if (ifp->if_baudrate == 0)
962 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
963 
964 	if (lla != NULL)
965 		if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla));
966 
967 	LIST_INIT(&ec->ec_multiaddrs);
968 	ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
969 	ifp->if_broadcastaddr = etherbroadcastaddr;
970 	bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header));
971 #ifdef MBUFTRACE
972 	strlcpy(ec->ec_tx_mowner.mo_name, ifp->if_xname,
973 	    sizeof(ec->ec_tx_mowner.mo_name));
974 	strlcpy(ec->ec_tx_mowner.mo_descr, "tx",
975 	    sizeof(ec->ec_tx_mowner.mo_descr));
976 	strlcpy(ec->ec_rx_mowner.mo_name, ifp->if_xname,
977 	    sizeof(ec->ec_rx_mowner.mo_name));
978 	strlcpy(ec->ec_rx_mowner.mo_descr, "rx",
979 	    sizeof(ec->ec_rx_mowner.mo_descr));
980 	MOWNER_ATTACH(&ec->ec_tx_mowner);
981 	MOWNER_ATTACH(&ec->ec_rx_mowner);
982 	ifp->if_mowner = &ec->ec_tx_mowner;
983 #endif
984 }
985 
986 void
987 ether_ifdetach(struct ifnet *ifp)
988 {
989 	struct ethercom *ec = (void *) ifp;
990 	struct ether_multi *enm;
991 
992 	/*
993 	 * Prevent further calls to ioctl (for example turning off
994 	 * promiscuous mode from the bridge code), which eventually can
995 	 * call if_init() which can cause panics because the interface
996 	 * is in the process of being detached. Return device not configured
997 	 * instead.
998 	 */
999 	ifp->if_ioctl = (int (*)(struct ifnet *, u_long, void *))enxio;
1000 
1001 #if NBRIDGE > 0
1002 	if (ifp->if_bridge)
1003 		bridge_ifdetach(ifp);
1004 #endif
1005 
1006 	bpf_detach(ifp);
1007 
1008 #if NVLAN > 0
1009 	if (ec->ec_nvlans)
1010 		vlan_ifdetach(ifp);
1011 #endif
1012 
1013 	mutex_enter(ec->ec_lock);
1014 	while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) {
1015 		LIST_REMOVE(enm, enm_list);
1016 		kmem_intr_free(enm, sizeof(*enm));
1017 		ec->ec_multicnt--;
1018 	}
1019 	mutex_exit(ec->ec_lock);
1020 
1021 	mutex_destroy(ec->ec_lock);
1022 
1023 	ifp->if_mowner = NULL;
1024 	MOWNER_DETACH(&ec->ec_rx_mowner);
1025 	MOWNER_DETACH(&ec->ec_tx_mowner);
1026 }
1027 
1028 #if 0
1029 /*
1030  * This is for reference.  We have a table-driven version
1031  * of the little-endian crc32 generator, which is faster
1032  * than the double-loop.
1033  */
1034 uint32_t
1035 ether_crc32_le(const uint8_t *buf, size_t len)
1036 {
1037 	uint32_t c, crc, carry;
1038 	size_t i, j;
1039 
1040 	crc = 0xffffffffU;	/* initial value */
1041 
1042 	for (i = 0; i < len; i++) {
1043 		c = buf[i];
1044 		for (j = 0; j < 8; j++) {
1045 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
1046 			crc >>= 1;
1047 			c >>= 1;
1048 			if (carry)
1049 				crc = (crc ^ ETHER_CRC_POLY_LE);
1050 		}
1051 	}
1052 
1053 	return (crc);
1054 }
1055 #else
1056 uint32_t
1057 ether_crc32_le(const uint8_t *buf, size_t len)
1058 {
1059 	static const uint32_t crctab[] = {
1060 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1061 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1062 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1063 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1064 	};
1065 	uint32_t crc;
1066 	size_t i;
1067 
1068 	crc = 0xffffffffU;	/* initial value */
1069 
1070 	for (i = 0; i < len; i++) {
1071 		crc ^= buf[i];
1072 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1073 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1074 	}
1075 
1076 	return (crc);
1077 }
1078 #endif
1079 
1080 uint32_t
1081 ether_crc32_be(const uint8_t *buf, size_t len)
1082 {
1083 	uint32_t c, crc, carry;
1084 	size_t i, j;
1085 
1086 	crc = 0xffffffffU;	/* initial value */
1087 
1088 	for (i = 0; i < len; i++) {
1089 		c = buf[i];
1090 		for (j = 0; j < 8; j++) {
1091 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
1092 			crc <<= 1;
1093 			c >>= 1;
1094 			if (carry)
1095 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1096 		}
1097 	}
1098 
1099 	return (crc);
1100 }
1101 
1102 #ifdef INET
1103 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] =
1104     { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 };
1105 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] =
1106     { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff };
1107 #endif
1108 #ifdef INET6
1109 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] =
1110     { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 };
1111 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] =
1112     { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff };
1113 #endif
1114 
1115 /*
1116  * ether_aton implementation, not using a static buffer.
1117  */
1118 int
1119 ether_aton_r(u_char *dest, size_t len, const char *str)
1120 {
1121         const u_char *cp = (const void *)str;
1122 	u_char *ep;
1123 
1124 #define atox(c)	(((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10))
1125 
1126 	if (len < ETHER_ADDR_LEN)
1127 		return ENOSPC;
1128 
1129 	ep = dest + ETHER_ADDR_LEN;
1130 
1131 	while (*cp) {
1132                 if (!isxdigit(*cp))
1133                         return EINVAL;
1134 		*dest = atox(*cp);
1135 		cp++;
1136                 if (isxdigit(*cp)) {
1137                         *dest = (*dest << 4) | atox(*cp);
1138 			dest++;
1139 			cp++;
1140                 } else
1141 			dest++;
1142 		if (dest == ep)
1143 			return *cp == '\0' ? 0 : ENAMETOOLONG;
1144 		switch (*cp) {
1145 		case ':':
1146 		case '-':
1147 		case '.':
1148 			cp++;
1149 			break;
1150 		}
1151         }
1152 	return ENOBUFS;
1153 }
1154 
1155 /*
1156  * Convert a sockaddr into an Ethernet address or range of Ethernet
1157  * addresses.
1158  */
1159 int
1160 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN],
1161     uint8_t addrhi[ETHER_ADDR_LEN])
1162 {
1163 #ifdef INET
1164 	const struct sockaddr_in *sin;
1165 #endif /* INET */
1166 #ifdef INET6
1167 	const struct sockaddr_in6 *sin6;
1168 #endif /* INET6 */
1169 
1170 	switch (sa->sa_family) {
1171 
1172 	case AF_UNSPEC:
1173 		memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN);
1174 		memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1175 		break;
1176 
1177 #ifdef INET
1178 	case AF_INET:
1179 		sin = satocsin(sa);
1180 		if (sin->sin_addr.s_addr == INADDR_ANY) {
1181 			/*
1182 			 * An IP address of INADDR_ANY means listen to
1183 			 * or stop listening to all of the Ethernet
1184 			 * multicast addresses used for IP.
1185 			 * (This is for the sake of IP multicast routers.)
1186 			 */
1187 			memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN);
1188 			memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN);
1189 		}
1190 		else {
1191 			ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo);
1192 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1193 		}
1194 		break;
1195 #endif
1196 #ifdef INET6
1197 	case AF_INET6:
1198 		sin6 = satocsin6(sa);
1199 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1200 			/*
1201 			 * An IP6 address of 0 means listen to or stop
1202 			 * listening to all of the Ethernet multicast
1203 			 * address used for IP6.
1204 			 * (This is used for multicast routers.)
1205 			 */
1206 			memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN);
1207 			memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN);
1208 		} else {
1209 			ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo);
1210 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
1211 		}
1212 		break;
1213 #endif
1214 
1215 	default:
1216 		return EAFNOSUPPORT;
1217 	}
1218 	return 0;
1219 }
1220 
1221 /*
1222  * Add an Ethernet multicast address or range of addresses to the list for a
1223  * given interface.
1224  */
1225 int
1226 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec)
1227 {
1228 	struct ether_multi *enm, *_enm;
1229 	u_char addrlo[ETHER_ADDR_LEN];
1230 	u_char addrhi[ETHER_ADDR_LEN];
1231 	int error = 0;
1232 
1233 	/* Allocate out of lock */
1234 	/* XXX still can be called in softint */
1235 	enm = kmem_intr_alloc(sizeof(*enm), KM_SLEEP);
1236 	if (enm == NULL)
1237 		return ENOBUFS;
1238 
1239 	mutex_enter(ec->ec_lock);
1240 	error = ether_multiaddr(sa, addrlo, addrhi);
1241 	if (error != 0)
1242 		goto out;
1243 
1244 	/*
1245 	 * Verify that we have valid Ethernet multicast addresses.
1246 	 */
1247 	if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) {
1248 		error = EINVAL;
1249 		goto out;
1250 	}
1251 	/*
1252 	 * See if the address range is already in the list.
1253 	 */
1254 	ETHER_LOOKUP_MULTI(addrlo, addrhi, ec, _enm);
1255 	if (_enm != NULL) {
1256 		/*
1257 		 * Found it; just increment the reference count.
1258 		 */
1259 		++_enm->enm_refcount;
1260 		error = 0;
1261 		goto out;
1262 	}
1263 	/*
1264 	 * Link a new multicast record into the interface's multicast list.
1265 	 */
1266 	memcpy(enm->enm_addrlo, addrlo, 6);
1267 	memcpy(enm->enm_addrhi, addrhi, 6);
1268 	enm->enm_refcount = 1;
1269 	LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list);
1270 	ec->ec_multicnt++;
1271 	/*
1272 	 * Return ENETRESET to inform the driver that the list has changed
1273 	 * and its reception filter should be adjusted accordingly.
1274 	 */
1275 	error = ENETRESET;
1276 	enm = NULL;
1277 out:
1278 	mutex_exit(ec->ec_lock);
1279 	if (enm != NULL)
1280 		kmem_intr_free(enm, sizeof(*enm));
1281 	return error;
1282 }
1283 
1284 /*
1285  * Delete a multicast address record.
1286  */
1287 int
1288 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec)
1289 {
1290 	struct ether_multi *enm;
1291 	u_char addrlo[ETHER_ADDR_LEN];
1292 	u_char addrhi[ETHER_ADDR_LEN];
1293 	int error;
1294 
1295 	mutex_enter(ec->ec_lock);
1296 	error = ether_multiaddr(sa, addrlo, addrhi);
1297 	if (error != 0)
1298 		goto error;
1299 
1300 	/*
1301 	 * Look ur the address in our list.
1302 	 */
1303 	ETHER_LOOKUP_MULTI(addrlo, addrhi, ec, enm);
1304 	if (enm == NULL) {
1305 		error = ENXIO;
1306 		goto error;
1307 	}
1308 	if (--enm->enm_refcount != 0) {
1309 		/*
1310 		 * Still some claims to this record.
1311 		 */
1312 		error = 0;
1313 		goto error;
1314 	}
1315 	/*
1316 	 * No remaining claims to this record; unlink and free it.
1317 	 */
1318 	LIST_REMOVE(enm, enm_list);
1319 	ec->ec_multicnt--;
1320 	mutex_exit(ec->ec_lock);
1321 
1322 	kmem_intr_free(enm, sizeof(*enm));
1323 	/*
1324 	 * Return ENETRESET to inform the driver that the list has changed
1325 	 * and its reception filter should be adjusted accordingly.
1326 	 */
1327 	return ENETRESET;
1328 error:
1329 	mutex_exit(ec->ec_lock);
1330 	return error;
1331 }
1332 
1333 void
1334 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb)
1335 {
1336 	ec->ec_ifflags_cb = cb;
1337 }
1338 
1339 /*
1340  * Common ioctls for Ethernet interfaces.  Note, we must be
1341  * called at splnet().
1342  */
1343 int
1344 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data)
1345 {
1346 	struct ethercom *ec = (void *) ifp;
1347 	struct eccapreq *eccr;
1348 	struct ifreq *ifr = (struct ifreq *)data;
1349 	struct if_laddrreq *iflr = data;
1350 	const struct sockaddr_dl *sdl;
1351 	static const uint8_t zero[ETHER_ADDR_LEN];
1352 	int error;
1353 
1354 	switch (cmd) {
1355 	case SIOCINITIFADDR:
1356 	    {
1357 		struct ifaddr *ifa = (struct ifaddr *)data;
1358 		if (ifa->ifa_addr->sa_family != AF_LINK
1359 		    && (ifp->if_flags & (IFF_UP|IFF_RUNNING)) !=
1360 		       (IFF_UP|IFF_RUNNING)) {
1361 			ifp->if_flags |= IFF_UP;
1362 			if ((error = (*ifp->if_init)(ifp)) != 0)
1363 				return error;
1364 		}
1365 #ifdef INET
1366 		if (ifa->ifa_addr->sa_family == AF_INET)
1367 			arp_ifinit(ifp, ifa);
1368 #endif /* INET */
1369 		return 0;
1370 	    }
1371 
1372 	case SIOCSIFMTU:
1373 	    {
1374 		int maxmtu;
1375 
1376 		if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU)
1377 			maxmtu = ETHERMTU_JUMBO;
1378 		else
1379 			maxmtu = ETHERMTU;
1380 
1381 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu)
1382 			return EINVAL;
1383 		else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET)
1384 			return error;
1385 		else if (ifp->if_flags & IFF_UP) {
1386 			/* Make sure the device notices the MTU change. */
1387 			return (*ifp->if_init)(ifp);
1388 		} else
1389 			return 0;
1390 	    }
1391 
1392 	case SIOCSIFFLAGS:
1393 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
1394 			return error;
1395 		switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
1396 		case IFF_RUNNING:
1397 			/*
1398 			 * If interface is marked down and it is running,
1399 			 * then stop and disable it.
1400 			 */
1401 			(*ifp->if_stop)(ifp, 1);
1402 			break;
1403 		case IFF_UP:
1404 			/*
1405 			 * If interface is marked up and it is stopped, then
1406 			 * start it.
1407 			 */
1408 			return (*ifp->if_init)(ifp);
1409 		case IFF_UP|IFF_RUNNING:
1410 			error = 0;
1411 			if (ec->ec_ifflags_cb == NULL ||
1412 			    (error = (*ec->ec_ifflags_cb)(ec)) == ENETRESET) {
1413 				/*
1414 				 * Reset the interface to pick up
1415 				 * changes in any other flags that
1416 				 * affect the hardware state.
1417 				 */
1418 				return (*ifp->if_init)(ifp);
1419 			} else
1420 				return error;
1421 		case 0:
1422 			break;
1423 		}
1424 		return 0;
1425 	case SIOCGETHERCAP:
1426 		eccr = (struct eccapreq *)data;
1427 		eccr->eccr_capabilities = ec->ec_capabilities;
1428 		eccr->eccr_capenable = ec->ec_capenable;
1429 		return 0;
1430 	case SIOCADDMULTI:
1431 		return ether_addmulti(ifreq_getaddr(cmd, ifr), ec);
1432 	case SIOCDELMULTI:
1433 		return ether_delmulti(ifreq_getaddr(cmd, ifr), ec);
1434 	case SIOCSIFMEDIA:
1435 	case SIOCGIFMEDIA:
1436 		if (ec->ec_mii == NULL)
1437 			return ENOTTY;
1438 		return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media, cmd);
1439 	case SIOCALIFADDR:
1440 		sdl = satocsdl(sstocsa(&iflr->addr));
1441 		if (sdl->sdl_family != AF_LINK)
1442 			;
1443 		else if (ETHER_IS_MULTICAST(CLLADDR(sdl)))
1444 			return EINVAL;
1445 		else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0)
1446 			return EINVAL;
1447 		/*FALLTHROUGH*/
1448 	default:
1449 		return ifioctl_common(ifp, cmd, data);
1450 	}
1451 	return 0;
1452 }
1453 
1454 /*
1455  * Enable/disable passing VLAN packets if the parent interface supports it.
1456  * Return:
1457  * 	 0: Ok
1458  *	-1: Parent interface does not support vlans
1459  *	>0: Error
1460  */
1461 int
1462 ether_enable_vlan_mtu(struct ifnet *ifp)
1463 {
1464 	int error;
1465 	struct ethercom *ec = (void *)ifp;
1466 
1467 	/* Parent does not support VLAN's */
1468 	if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0)
1469 		return -1;
1470 
1471 	/*
1472 	 * Parent supports the VLAN_MTU capability,
1473 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames;
1474 	 * enable it.
1475 	 */
1476 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1477 
1478 	/* Interface is down, defer for later */
1479 	if ((ifp->if_flags & IFF_UP) == 0)
1480 		return 0;
1481 
1482 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1483 		return 0;
1484 
1485 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1486 	return error;
1487 }
1488 
1489 int
1490 ether_disable_vlan_mtu(struct ifnet *ifp)
1491 {
1492 	int error;
1493 	struct ethercom *ec = (void *)ifp;
1494 
1495 	/* We still have VLAN's, defer for later */
1496 	if (ec->ec_nvlans != 0)
1497 		return 0;
1498 
1499 	/* Parent does not support VLAB's, nothing to do. */
1500 	if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0)
1501 		return -1;
1502 
1503 	/*
1504 	 * Disable Tx/Rx of VLAN-sized frames.
1505 	 */
1506 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
1507 
1508 	/* Interface is down, defer for later */
1509 	if ((ifp->if_flags & IFF_UP) == 0)
1510 		return 0;
1511 
1512 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
1513 		return 0;
1514 
1515 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
1516 	return error;
1517 }
1518 
1519 static int
1520 ether_multicast_sysctl(SYSCTLFN_ARGS)
1521 {
1522 	struct ether_multi *enm;
1523 	struct ifnet *ifp;
1524 	struct ethercom *ec;
1525 	int error = 0;
1526 	size_t written;
1527 	struct psref psref;
1528 	int bound;
1529 	unsigned int multicnt;
1530 	struct ether_multi_sysctl *addrs;
1531 	int i;
1532 
1533 	if (namelen != 1)
1534 		return EINVAL;
1535 
1536 	bound = curlwp_bind();
1537 	ifp = if_get_byindex(name[0], &psref);
1538 	if (ifp == NULL) {
1539 		error = ENODEV;
1540 		goto out;
1541 	}
1542 	if (ifp->if_type != IFT_ETHER) {
1543 		if_put(ifp, &psref);
1544 		*oldlenp = 0;
1545 		goto out;
1546 	}
1547 	ec = (struct ethercom *)ifp;
1548 
1549 	if (oldp == NULL) {
1550 		if_put(ifp, &psref);
1551 		*oldlenp = ec->ec_multicnt * sizeof(*addrs);
1552 		goto out;
1553 	}
1554 
1555 	/*
1556 	 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which
1557 	 * is sleepable, with holding it. Copy data to a local buffer first
1558 	 * with holding it and then call sysctl_copyout without holding it.
1559 	 */
1560 retry:
1561 	multicnt = ec->ec_multicnt;
1562 	addrs = kmem_alloc(sizeof(*addrs) * multicnt, KM_SLEEP);
1563 
1564 	mutex_enter(ec->ec_lock);
1565 	if (multicnt < ec->ec_multicnt) {
1566 		/* The number of multicast addresses have increased */
1567 		mutex_exit(ec->ec_lock);
1568 		kmem_free(addrs, sizeof(*addrs) * multicnt);
1569 		goto retry;
1570 	}
1571 
1572 	i = 0;
1573 	LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) {
1574 		struct ether_multi_sysctl *addr = &addrs[i];
1575 		addr->enm_refcount = enm->enm_refcount;
1576 		memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN);
1577 		memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN);
1578 		i++;
1579 	}
1580 	mutex_exit(ec->ec_lock);
1581 
1582 	error = 0;
1583 	written = 0;
1584 	for (i = 0; i < multicnt; i++) {
1585 		struct ether_multi_sysctl *addr = &addrs[i];
1586 
1587 		if (written + sizeof(*addr) > *oldlenp)
1588 			break;
1589 		error = sysctl_copyout(l, addr, oldp, sizeof(*addr));
1590 		if (error)
1591 			break;
1592 		written += sizeof(*addr);
1593 		oldp = (char *)oldp + sizeof(*addr);
1594 	}
1595 	kmem_free(addrs, sizeof(*addrs) * multicnt);
1596 
1597 	if_put(ifp, &psref);
1598 
1599 	*oldlenp = written;
1600 out:
1601 	curlwp_bindx(bound);
1602 	return error;
1603 }
1604 
1605 static void
1606 ether_sysctl_setup(struct sysctllog **clog)
1607 {
1608 	const struct sysctlnode *rnode = NULL;
1609 
1610 	sysctl_createv(clog, 0, NULL, &rnode,
1611 		       CTLFLAG_PERMANENT,
1612 		       CTLTYPE_NODE, "ether",
1613 		       SYSCTL_DESCR("Ethernet-specific information"),
1614 		       NULL, 0, NULL, 0,
1615 		       CTL_NET, CTL_CREATE, CTL_EOL);
1616 
1617 	sysctl_createv(clog, 0, &rnode, NULL,
1618 		       CTLFLAG_PERMANENT,
1619 		       CTLTYPE_NODE, "multicast",
1620 		       SYSCTL_DESCR("multicast addresses"),
1621 		       ether_multicast_sysctl, 0, NULL, 0,
1622 		       CTL_CREATE, CTL_EOL);
1623 }
1624 
1625 void
1626 etherinit(void)
1627 {
1628 
1629 	mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET);
1630 	ether_sysctl_setup(NULL);
1631 }
1632