Home | History | Annotate | Line # | Download | only in net
if_ethersubr.c revision 1.218
      1 /*	$NetBSD: if_ethersubr.c,v 1.218 2016/04/15 01:31:29 ozaki-r Exp $	*/
      2 
      3 /*
      4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
      5  * All rights reserved.
      6  *
      7  * Redistribution and use in source and binary forms, with or without
      8  * modification, are permitted provided that the following conditions
      9  * are met:
     10  * 1. Redistributions of source code must retain the above copyright
     11  *    notice, this list of conditions and the following disclaimer.
     12  * 2. Redistributions in binary form must reproduce the above copyright
     13  *    notice, this list of conditions and the following disclaimer in the
     14  *    documentation and/or other materials provided with the distribution.
     15  * 3. Neither the name of the project nor the names of its contributors
     16  *    may be used to endorse or promote products derived from this software
     17  *    without specific prior written permission.
     18  *
     19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
     20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
     23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     29  * SUCH DAMAGE.
     30  */
     31 
     32 /*
     33  * Copyright (c) 1982, 1989, 1993
     34  *	The Regents of the University of California.  All rights reserved.
     35  *
     36  * Redistribution and use in source and binary forms, with or without
     37  * modification, are permitted provided that the following conditions
     38  * are met:
     39  * 1. Redistributions of source code must retain the above copyright
     40  *    notice, this list of conditions and the following disclaimer.
     41  * 2. Redistributions in binary form must reproduce the above copyright
     42  *    notice, this list of conditions and the following disclaimer in the
     43  *    documentation and/or other materials provided with the distribution.
     44  * 3. Neither the name of the University nor the names of its contributors
     45  *    may be used to endorse or promote products derived from this software
     46  *    without specific prior written permission.
     47  *
     48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     58  * SUCH DAMAGE.
     59  *
     60  *	@(#)if_ethersubr.c	8.2 (Berkeley) 4/4/96
     61  */
     62 
     63 #include <sys/cdefs.h>
     64 __KERNEL_RCSID(0, "$NetBSD: if_ethersubr.c,v 1.218 2016/04/15 01:31:29 ozaki-r Exp $");
     65 
     66 #ifdef _KERNEL_OPT
     67 #include "opt_inet.h"
     68 #include "opt_atalk.h"
     69 #include "opt_mbuftrace.h"
     70 #include "opt_mpls.h"
     71 #include "opt_gateway.h"
     72 #include "opt_pppoe.h"
     73 #include "opt_net_mpsafe.h"
     74 #endif
     75 
     76 #include "vlan.h"
     77 #include "pppoe.h"
     78 #include "bridge.h"
     79 #include "arp.h"
     80 #include "agr.h"
     81 
     82 #include <sys/sysctl.h>
     83 #include <sys/malloc.h>
     84 #include <sys/mbuf.h>
     85 #include <sys/mutex.h>
     86 #include <sys/ioctl.h>
     87 #include <sys/errno.h>
     88 #include <sys/device.h>
     89 #include <sys/rnd.h>
     90 #include <sys/rndsource.h>
     91 #include <sys/cpu.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 *, 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 	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 	ALTQ_DECL(struct altq_pktattr pktattr;)
    201 #ifdef INET
    202 	struct arphdr *ah;
    203 #endif /* INET */
    204 #ifdef NETATALK
    205 	struct at_ifaddr *aa;
    206 #endif /* NETATALK */
    207 
    208 #ifndef NET_MPSAFE
    209 	KASSERT(KERNEL_LOCKED_P());
    210 #endif
    211 
    212 #ifdef MBUFTRACE
    213 	m_claimm(m, ifp->if_mowner);
    214 #endif
    215 
    216 #if NCARP > 0
    217 	if (ifp->if_type == IFT_CARP) {
    218 		struct ifaddr *ifa;
    219 
    220 		/* loop back if this is going to the carp interface */
    221 		if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP &&
    222 		    (ifa = ifa_ifwithaddr(dst)) != NULL &&
    223 		    ifa->ifa_ifp == ifp0)
    224 			return looutput(ifp0, m, dst, rt);
    225 
    226 		ifp = ifp->if_carpdev;
    227 		/* ac = (struct arpcom *)ifp; */
    228 
    229 		if ((ifp0->if_flags & (IFF_UP|IFF_RUNNING)) !=
    230 		    (IFF_UP|IFF_RUNNING))
    231 			senderr(ENETDOWN);
    232 	}
    233 #endif /* NCARP > 0 */
    234 
    235 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
    236 		senderr(ENETDOWN);
    237 
    238 	switch (dst->sa_family) {
    239 
    240 #ifdef INET
    241 	case AF_INET:
    242 		if (m->m_flags & M_BCAST)
    243 			(void)memcpy(edst, etherbroadcastaddr, sizeof(edst));
    244 		else if (m->m_flags & M_MCAST)
    245 			ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst);
    246 		else if ((error = arpresolve(ifp, rt, m, dst, edst,
    247 		    sizeof(edst))) != 0)
    248 			return error == EWOULDBLOCK ? 0 : error;
    249 		/* If broadcasting on a simplex interface, loopback a copy */
    250 		if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
    251 			mcopy = m_copy(m, 0, (int)M_COPYALL);
    252 		etype = htons(ETHERTYPE_IP);
    253 		break;
    254 
    255 	case AF_ARP:
    256 		ah = mtod(m, struct arphdr *);
    257 		if (m->m_flags & M_BCAST)
    258 			(void)memcpy(edst, etherbroadcastaddr, sizeof(edst));
    259 		else {
    260 			void *tha = ar_tha(ah);
    261 
    262 			if (tha == NULL) {
    263 				/* fake with ARPHDR_IEEE1394 */
    264 				return 0;
    265 			}
    266 			memcpy(edst, tha, sizeof(edst));
    267 		}
    268 
    269 		ah->ar_hrd = htons(ARPHRD_ETHER);
    270 
    271 		switch (ntohs(ah->ar_op)) {
    272 		case ARPOP_REVREQUEST:
    273 		case ARPOP_REVREPLY:
    274 			etype = htons(ETHERTYPE_REVARP);
    275 			break;
    276 
    277 		case ARPOP_REQUEST:
    278 		case ARPOP_REPLY:
    279 		default:
    280 			etype = htons(ETHERTYPE_ARP);
    281 		}
    282 
    283 		break;
    284 #endif
    285 #ifdef INET6
    286 	case AF_INET6:
    287 		if (!nd6_storelladdr(ifp, rt, m, dst, edst, sizeof(edst))){
    288 			/* something bad happened */
    289 			return (0);
    290 		}
    291 		etype = htons(ETHERTYPE_IPV6);
    292 		break;
    293 #endif
    294 #ifdef NETATALK
    295     case AF_APPLETALK:
    296 		if (aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) {
    297 #ifdef NETATALKDEBUG
    298 			printf("aarpresolv failed\n");
    299 #endif /* NETATALKDEBUG */
    300 			return (0);
    301 		}
    302 		/*
    303 		 * ifaddr is the first thing in at_ifaddr
    304 		 */
    305 		aa = (struct at_ifaddr *) at_ifawithnet(
    306 		    (const struct sockaddr_at *)dst, ifp);
    307 		if (aa == NULL)
    308 		    goto bad;
    309 
    310 		/*
    311 		 * In the phase 2 case, we need to prepend an mbuf for the
    312 		 * llc header.  Since we must preserve the value of m,
    313 		 * which is passed to us by value, we m_copy() the first
    314 		 * mbuf, and use it for our llc header.
    315 		 */
    316 		if (aa->aa_flags & AFA_PHASE2) {
    317 			struct llc llc;
    318 
    319 			M_PREPEND(m, sizeof(struct llc), M_DONTWAIT);
    320 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
    321 			llc.llc_control = LLC_UI;
    322 			memcpy(llc.llc_snap_org_code, at_org_code,
    323 			    sizeof(llc.llc_snap_org_code));
    324 			llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK);
    325 			memcpy(mtod(m, void *), &llc, sizeof(struct llc));
    326 		} else {
    327 			etype = htons(ETHERTYPE_ATALK);
    328 		}
    329 		break;
    330 #endif /* NETATALK */
    331 	case pseudo_AF_HDRCMPLT:
    332 		hdrcmplt = 1;
    333 		memcpy(esrc,
    334 		    ((const struct ether_header *)dst->sa_data)->ether_shost,
    335 		    sizeof(esrc));
    336 		/* FALLTHROUGH */
    337 
    338 	case AF_UNSPEC:
    339  		memcpy(edst,
    340 		    ((const struct ether_header *)dst->sa_data)->ether_dhost,
    341 		    sizeof(edst));
    342 		/* AF_UNSPEC doesn't swap the byte order of the ether_type. */
    343 		etype = ((const struct ether_header *)dst->sa_data)->ether_type;
    344 		break;
    345 
    346 	default:
    347 		printf("%s: can't handle af%d\n", ifp->if_xname,
    348 			dst->sa_family);
    349 		senderr(EAFNOSUPPORT);
    350 	}
    351 
    352 #ifdef MPLS
    353 	{
    354 		struct m_tag *mtag;
    355 		mtag = m_tag_find(m, PACKET_TAG_MPLS, NULL);
    356 		if (mtag != NULL) {
    357 			/* Having the tag itself indicates it's MPLS */
    358 			etype = htons(ETHERTYPE_MPLS);
    359 			m_tag_delete(m, mtag);
    360 		}
    361 	}
    362 #endif
    363 
    364 	if (mcopy)
    365 		(void)looutput(ifp, mcopy, dst, rt);
    366 
    367 	/* If no ether type is set, this must be a 802.2 formatted packet.
    368 	 */
    369 	if (etype == 0)
    370 		etype = htons(m->m_pkthdr.len);
    371 	/*
    372 	 * Add local net header.  If no space in first mbuf,
    373 	 * allocate another.
    374 	 */
    375 	M_PREPEND(m, sizeof (struct ether_header), M_DONTWAIT);
    376 	if (m == 0)
    377 		senderr(ENOBUFS);
    378 	eh = mtod(m, struct ether_header *);
    379 	/* Note: etype is already in network byte order. */
    380 	(void)memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
    381  	memcpy(eh->ether_dhost, edst, sizeof(edst));
    382 	if (hdrcmplt)
    383 		memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost));
    384 	else
    385 	 	memcpy(eh->ether_shost, CLLADDR(ifp->if_sadl),
    386 		    sizeof(eh->ether_shost));
    387 
    388 #if NCARP > 0
    389 	if (ifp0 != ifp && ifp0->if_type == IFT_CARP) {
    390 	 	memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl),
    391 		    sizeof(eh->ether_shost));
    392 	}
    393 #endif /* NCARP > 0 */
    394 
    395 	if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0)
    396 		return (error);
    397 	if (m == NULL)
    398 		return (0);
    399 
    400 #if NBRIDGE > 0
    401 	/*
    402 	 * Bridges require special output handling.
    403 	 */
    404 	if (ifp->if_bridge)
    405 		return (bridge_output(ifp, m, NULL, NULL));
    406 #endif
    407 
    408 #if NCARP > 0
    409 	if (ifp != ifp0)
    410 		ifp0->if_obytes += m->m_pkthdr.len + ETHER_HDR_LEN;
    411 #endif /* NCARP > 0 */
    412 
    413 #ifdef ALTQ
    414 	/*
    415 	 * If ALTQ is enabled on the parent interface, do
    416 	 * classification; the queueing discipline might not
    417 	 * require classification, but might require the
    418 	 * address family/header pointer in the pktattr.
    419 	 */
    420 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
    421 		altq_etherclassify(&ifp->if_snd, m, &pktattr);
    422 #endif
    423 	return ifq_enqueue(ifp, m ALTQ_COMMA ALTQ_DECL(&pktattr));
    424 
    425 bad:
    426 	if (m)
    427 		m_freem(m);
    428 	return (error);
    429 }
    430 
    431 #ifdef ALTQ
    432 /*
    433  * This routine is a slight hack to allow a packet to be classified
    434  * if the Ethernet headers are present.  It will go away when ALTQ's
    435  * classification engine understands link headers.
    436  */
    437 void
    438 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
    439     struct altq_pktattr *pktattr)
    440 {
    441 	struct ether_header *eh;
    442 	uint16_t ether_type;
    443 	int hlen, af, hdrsize;
    444 	void *hdr;
    445 
    446 	hlen = ETHER_HDR_LEN;
    447 	eh = mtod(m, struct ether_header *);
    448 
    449 	ether_type = htons(eh->ether_type);
    450 
    451 	if (ether_type < ETHERMTU) {
    452 		/* LLC/SNAP */
    453 		struct llc *llc = (struct llc *)(eh + 1);
    454 		hlen += 8;
    455 
    456 		if (m->m_len < hlen ||
    457 		    llc->llc_dsap != LLC_SNAP_LSAP ||
    458 		    llc->llc_ssap != LLC_SNAP_LSAP ||
    459 		    llc->llc_control != LLC_UI) {
    460 			/* Not SNAP. */
    461 			goto bad;
    462 		}
    463 
    464 		ether_type = htons(llc->llc_un.type_snap.ether_type);
    465 	}
    466 
    467 	switch (ether_type) {
    468 	case ETHERTYPE_IP:
    469 		af = AF_INET;
    470 		hdrsize = 20;		/* sizeof(struct ip) */
    471 		break;
    472 
    473 	case ETHERTYPE_IPV6:
    474 		af = AF_INET6;
    475 		hdrsize = 40;		/* sizeof(struct ip6_hdr) */
    476 		break;
    477 
    478 	default:
    479 		af = AF_UNSPEC;
    480 		hdrsize = 0;
    481 		break;
    482 	}
    483 
    484 	while (m->m_len <= hlen) {
    485 		hlen -= m->m_len;
    486 		m = m->m_next;
    487 	}
    488 	if (m->m_len < (hlen + hdrsize)) {
    489 		/*
    490 		 * protocol header not in a single mbuf.
    491 		 * We can't cope with this situation right
    492 		 * now (but it shouldn't ever happen, really, anyhow).
    493 		 */
    494 #ifdef DEBUG
    495 		printf("altq_etherclassify: headers span multiple mbufs: "
    496 		    "%d < %d\n", m->m_len, (hlen + hdrsize));
    497 #endif
    498 		goto bad;
    499 	}
    500 
    501 	m->m_data += hlen;
    502 	m->m_len -= hlen;
    503 
    504 	hdr = mtod(m, void *);
    505 
    506 	if (ALTQ_NEEDS_CLASSIFY(ifq))
    507 		pktattr->pattr_class =
    508 		    (*ifq->altq_classify)(ifq->altq_clfier, m, af);
    509 	pktattr->pattr_af = af;
    510 	pktattr->pattr_hdr = hdr;
    511 
    512 	m->m_data -= hlen;
    513 	m->m_len += hlen;
    514 
    515 	return;
    516 
    517  bad:
    518 	pktattr->pattr_class = NULL;
    519 	pktattr->pattr_hdr = NULL;
    520 	pktattr->pattr_af = AF_UNSPEC;
    521 }
    522 #endif /* ALTQ */
    523 
    524 /*
    525  * Process a received Ethernet packet;
    526  * the packet is in the mbuf chain m with
    527  * the ether header.
    528  */
    529 void
    530 ether_input(struct ifnet *ifp, struct mbuf *m)
    531 {
    532 	struct ethercom *ec = (struct ethercom *) ifp;
    533 	pktqueue_t *pktq = NULL;
    534 	struct ifqueue *inq = NULL;
    535 	uint16_t etype;
    536 	struct ether_header *eh;
    537 	size_t ehlen;
    538 	static int earlypkts;
    539 	int isr = 0;
    540 #if defined (LLC) || defined(NETATALK)
    541 	struct llc *l;
    542 #endif
    543 
    544 	KASSERT(!cpu_intr_p());
    545 
    546 	if ((ifp->if_flags & IFF_UP) == 0) {
    547 		m_freem(m);
    548 		return;
    549 	}
    550 
    551 #ifdef MBUFTRACE
    552 	m_claimm(m, &ec->ec_rx_mowner);
    553 #endif
    554 	eh = mtod(m, struct ether_header *);
    555 	etype = ntohs(eh->ether_type);
    556 	ehlen = sizeof(*eh);
    557 
    558 	if(__predict_false(earlypkts < 100 || !rnd_initial_entropy)) {
    559 		rnd_add_data(NULL, eh, ehlen, 0);
    560 		earlypkts++;
    561 	}
    562 
    563 	/*
    564 	 * Determine if the packet is within its size limits.
    565 	 */
    566 	if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len >
    567 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
    568 		mutex_enter(&bigpktpps_lock);
    569 		if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count,
    570 			    bigpktppslim)) {
    571 			printf("%s: discarding oversize frame (len=%d)\n",
    572 			    ifp->if_xname, m->m_pkthdr.len);
    573 		}
    574 		mutex_exit(&bigpktpps_lock);
    575 		m_freem(m);
    576 		return;
    577 	}
    578 
    579 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
    580 		/*
    581 		 * If this is not a simplex interface, drop the packet
    582 		 * if it came from us.
    583 		 */
    584 		if ((ifp->if_flags & IFF_SIMPLEX) == 0 &&
    585 		    memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost,
    586 		    ETHER_ADDR_LEN) == 0) {
    587 			m_freem(m);
    588 			return;
    589 		}
    590 
    591 		if (memcmp(etherbroadcastaddr,
    592 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
    593 			m->m_flags |= M_BCAST;
    594 		else
    595 			m->m_flags |= M_MCAST;
    596 		ifp->if_imcasts++;
    597 	}
    598 
    599 	/* If the CRC is still on the packet, trim it off. */
    600 	if (m->m_flags & M_HASFCS) {
    601 		m_adj(m, -ETHER_CRC_LEN);
    602 		m->m_flags &= ~M_HASFCS;
    603 	}
    604 
    605 	ifp->if_ibytes += m->m_pkthdr.len;
    606 
    607 #if NCARP > 0
    608 	if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) {
    609 		/*
    610 		 * clear M_PROMISC, in case the packets comes from a
    611 		 * vlan
    612 		 */
    613 		m->m_flags &= ~M_PROMISC;
    614 		if (carp_input(m, (uint8_t *)&eh->ether_shost,
    615 		    (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0)
    616 			return;
    617 	}
    618 #endif /* NCARP > 0 */
    619 	if ((m->m_flags & (M_BCAST|M_MCAST|M_PROMISC)) == 0 &&
    620 	    (ifp->if_flags & IFF_PROMISC) != 0 &&
    621 	    memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost,
    622 		   ETHER_ADDR_LEN) != 0) {
    623 		m->m_flags |= M_PROMISC;
    624 	}
    625 
    626 	if ((m->m_flags & M_PROMISC) == 0) {
    627 		if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
    628 			return;
    629 		if (m == NULL)
    630 			return;
    631 
    632 		eh = mtod(m, struct ether_header *);
    633 		etype = ntohs(eh->ether_type);
    634 		ehlen = sizeof(*eh);
    635 	}
    636 
    637 #if NAGR > 0
    638 	if (ifp->if_agrprivate &&
    639 	    __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) {
    640 		m->m_flags &= ~M_PROMISC;
    641 		agr_input(ifp, m);
    642 		return;
    643 	}
    644 #endif /* NAGR > 0 */
    645 
    646 	/*
    647 	 * If VLANs are configured on the interface, check to
    648 	 * see if the device performed the decapsulation and
    649 	 * provided us with the tag.
    650 	 */
    651 	if (ec->ec_nvlans && m_tag_find(m, PACKET_TAG_VLAN, NULL) != NULL) {
    652 #if NVLAN > 0
    653 		/*
    654 		 * vlan_input() will either recursively call ether_input()
    655 		 * or drop the packet.
    656 		 */
    657 		vlan_input(ifp, m);
    658 #else
    659 		m_freem(m);
    660 #endif
    661 		return;
    662 	}
    663 
    664 	/*
    665 	 * Handle protocols that expect to have the Ethernet header
    666 	 * (and possibly FCS) intact.
    667 	 */
    668 	switch (etype) {
    669 	case ETHERTYPE_VLAN: {
    670 		struct ether_vlan_header *evl = (void *)eh;
    671 		/*
    672 		 * If there is a tag of 0, then the VLAN header was probably
    673 		 * just being used to store the priority.  Extract the ether
    674 		 * type, and if IP or IPV6, let them deal with it.
    675 		 */
    676 		if (m->m_len <= sizeof(*evl)
    677 		    && EVL_VLANOFTAG(evl->evl_tag) == 0) {
    678 			etype = ntohs(evl->evl_proto);
    679 			ehlen = sizeof(*evl);
    680 			if ((m->m_flags & M_PROMISC) == 0
    681 			    && (etype == ETHERTYPE_IP
    682 				|| etype == ETHERTYPE_IPV6))
    683 				break;
    684 		}
    685 #if NVLAN > 0
    686 		/*
    687 		 * vlan_input() will either recursively call ether_input()
    688 		 * or drop the packet.
    689 		 */
    690 		if (((struct ethercom *)ifp)->ec_nvlans != 0)
    691 			vlan_input(ifp, m);
    692 		else
    693 #endif /* NVLAN > 0 */
    694 			m_freem(m);
    695 		return;
    696 	}
    697 #if NPPPOE > 0
    698 	case ETHERTYPE_PPPOEDISC:
    699 		pppoedisc_input(ifp, m);
    700 		return;
    701 	case ETHERTYPE_PPPOE:
    702 		pppoe_input(ifp, m);
    703 		return;
    704 #endif /* NPPPOE > 0 */
    705 	case ETHERTYPE_SLOWPROTOCOLS: {
    706 		uint8_t subtype;
    707 
    708 #if defined(DIAGNOSTIC)
    709 		if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype)) {
    710 			panic("ether_input: too short slow protocol packet");
    711 		}
    712 #endif
    713 		m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype);
    714 		switch (subtype) {
    715 #if NAGR > 0
    716 		case SLOWPROTOCOLS_SUBTYPE_LACP:
    717 			if (ifp->if_agrprivate) {
    718 				ieee8023ad_lacp_input(ifp, m);
    719 				return;
    720 			}
    721 			break;
    722 
    723 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
    724 			if (ifp->if_agrprivate) {
    725 				ieee8023ad_marker_input(ifp, m);
    726 				return;
    727 			}
    728 			break;
    729 #endif /* NAGR > 0 */
    730 		default:
    731 			if (subtype == 0 || subtype > 10) {
    732 				/* illegal value */
    733 				m_freem(m);
    734 				return;
    735 			}
    736 			/* unknown subtype */
    737 			break;
    738 		}
    739 		/* FALLTHROUGH */
    740 	}
    741 	default:
    742 		if (m->m_flags & M_PROMISC) {
    743 			m_freem(m);
    744 			return;
    745 		}
    746 	}
    747 
    748 	/* If the CRC is still on the packet, trim it off. */
    749 	if (m->m_flags & M_HASFCS) {
    750 		m_adj(m, -ETHER_CRC_LEN);
    751 		m->m_flags &= ~M_HASFCS;
    752 	}
    753 
    754 	if (etype > ETHERMTU + sizeof (struct ether_header)) {
    755 		/* Strip off the Ethernet header. */
    756 		m_adj(m, ehlen);
    757 
    758 		switch (etype) {
    759 #ifdef INET
    760 		case ETHERTYPE_IP:
    761 #ifdef GATEWAY
    762 			if (ipflow_fastforward(m))
    763 				return;
    764 #endif
    765 			pktq = ip_pktq;
    766 			break;
    767 
    768 		case ETHERTYPE_ARP:
    769 			isr = NETISR_ARP;
    770 			inq = &arpintrq;
    771 			break;
    772 
    773 		case ETHERTYPE_REVARP:
    774 			revarpinput(m);	/* XXX queue? */
    775 			return;
    776 #endif
    777 #ifdef INET6
    778 		case ETHERTYPE_IPV6:
    779 			if (__predict_false(!in6_present)) {
    780 				m_freem(m);
    781 				return;
    782 			}
    783 #ifdef GATEWAY
    784 			if (ip6flow_fastforward(&m))
    785 				return;
    786 #endif
    787 			pktq = ip6_pktq;
    788 			break;
    789 #endif
    790 #ifdef NETATALK
    791 		case ETHERTYPE_ATALK:
    792 			isr = NETISR_ATALK;
    793 			inq = &atintrq1;
    794 			break;
    795 		case ETHERTYPE_AARP:
    796 			/* probably this should be done with a NETISR as well */
    797 			aarpinput(ifp, m); /* XXX */
    798 			return;
    799 #endif /* NETATALK */
    800 #ifdef MPLS
    801 		case ETHERTYPE_MPLS:
    802 			isr = NETISR_MPLS;
    803 			inq = &mplsintrq;
    804 			break;
    805 #endif
    806 		default:
    807 			m_freem(m);
    808 			return;
    809 		}
    810 	} else {
    811 #if defined (LLC) || defined (NETATALK)
    812 		l = (struct llc *)(eh+1);
    813 		switch (l->llc_dsap) {
    814 #ifdef NETATALK
    815 		case LLC_SNAP_LSAP:
    816 			switch (l->llc_control) {
    817 			case LLC_UI:
    818 				if (l->llc_ssap != LLC_SNAP_LSAP) {
    819 					goto dropanyway;
    820 				}
    821 
    822 				if (memcmp(&(l->llc_snap_org_code)[0],
    823 				    at_org_code, sizeof(at_org_code)) == 0 &&
    824 				    ntohs(l->llc_snap_ether_type) ==
    825 				    ETHERTYPE_ATALK) {
    826 					inq = &atintrq2;
    827 					m_adj(m, sizeof(struct ether_header)
    828 					    + sizeof(struct llc));
    829 					isr = NETISR_ATALK;
    830 					break;
    831 				}
    832 
    833 				if (memcmp(&(l->llc_snap_org_code)[0],
    834 				    aarp_org_code,
    835 				    sizeof(aarp_org_code)) == 0 &&
    836 				    ntohs(l->llc_snap_ether_type) ==
    837 				    ETHERTYPE_AARP) {
    838 					m_adj( m, sizeof(struct ether_header)
    839 					    + sizeof(struct llc));
    840 					aarpinput(ifp, m); /* XXX */
    841 				    return;
    842 				}
    843 
    844 			default:
    845 				goto dropanyway;
    846 			}
    847 			break;
    848 		dropanyway:
    849 #endif
    850 		default:
    851 			m_freem(m);
    852 			return;
    853 		}
    854 #else /* ISO || LLC || NETATALK*/
    855 		m_freem(m);
    856 		return;
    857 #endif /* ISO || LLC || NETATALK*/
    858 	}
    859 
    860 	if (__predict_true(pktq)) {
    861 		const uint32_t h = pktq_rps_hash(m);
    862 		if (__predict_false(!pktq_enqueue(pktq, m, h))) {
    863 			m_freem(m);
    864 		}
    865 		return;
    866 	}
    867 
    868 	if (__predict_false(!inq)) {
    869 		/* Should not happen. */
    870 		m_freem(m);
    871 		return;
    872 	}
    873 	if (IF_QFULL(inq)) {
    874 		IF_DROP(inq);
    875 		m_freem(m);
    876 	} else {
    877 		IF_ENQUEUE(inq, m);
    878 		schednetisr(isr);
    879 	}
    880 }
    881 
    882 /*
    883  * Convert Ethernet address to printable (loggable) representation.
    884  */
    885 char *
    886 ether_sprintf(const u_char *ap)
    887 {
    888 	static char etherbuf[3 * ETHER_ADDR_LEN];
    889 	return ether_snprintf(etherbuf, sizeof(etherbuf), ap);
    890 }
    891 
    892 char *
    893 ether_snprintf(char *buf, size_t len, const u_char *ap)
    894 {
    895 	char *cp = buf;
    896 	size_t i;
    897 
    898 	for (i = 0; i < len / 3; i++) {
    899 		*cp++ = hexdigits[*ap >> 4];
    900 		*cp++ = hexdigits[*ap++ & 0xf];
    901 		*cp++ = ':';
    902 	}
    903 	*--cp = '\0';
    904 	return buf;
    905 }
    906 
    907 /*
    908  * Perform common duties while attaching to interface list
    909  */
    910 void
    911 ether_ifattach(struct ifnet *ifp, const uint8_t *lla)
    912 {
    913 	struct ethercom *ec = (struct ethercom *)ifp;
    914 
    915 	ifp->if_type = IFT_ETHER;
    916 	ifp->if_hdrlen = ETHER_HDR_LEN;
    917 	ifp->if_dlt = DLT_EN10MB;
    918 	ifp->if_mtu = ETHERMTU;
    919 	ifp->if_output = ether_output;
    920 	ifp->_if_input = ether_input;
    921 	if (ifp->if_baudrate == 0)
    922 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
    923 
    924 	if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla));
    925 
    926 	LIST_INIT(&ec->ec_multiaddrs);
    927 	ifp->if_broadcastaddr = etherbroadcastaddr;
    928 	bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header));
    929 #ifdef MBUFTRACE
    930 	strlcpy(ec->ec_tx_mowner.mo_name, ifp->if_xname,
    931 	    sizeof(ec->ec_tx_mowner.mo_name));
    932 	strlcpy(ec->ec_tx_mowner.mo_descr, "tx",
    933 	    sizeof(ec->ec_tx_mowner.mo_descr));
    934 	strlcpy(ec->ec_rx_mowner.mo_name, ifp->if_xname,
    935 	    sizeof(ec->ec_rx_mowner.mo_name));
    936 	strlcpy(ec->ec_rx_mowner.mo_descr, "rx",
    937 	    sizeof(ec->ec_rx_mowner.mo_descr));
    938 	MOWNER_ATTACH(&ec->ec_tx_mowner);
    939 	MOWNER_ATTACH(&ec->ec_rx_mowner);
    940 	ifp->if_mowner = &ec->ec_tx_mowner;
    941 #endif
    942 }
    943 
    944 void
    945 ether_ifdetach(struct ifnet *ifp)
    946 {
    947 	struct ethercom *ec = (void *) ifp;
    948 	struct ether_multi *enm;
    949 	int s;
    950 
    951 	/*
    952 	 * Prevent further calls to ioctl (for example turning off
    953 	 * promiscuous mode from the bridge code), which eventually can
    954 	 * call if_init() which can cause panics because the interface
    955 	 * is in the process of being detached. Return device not configured
    956 	 * instead.
    957 	 */
    958 	ifp->if_ioctl = (int (*)(struct ifnet *, u_long, void *))enxio;
    959 
    960 #if NBRIDGE > 0
    961 	if (ifp->if_bridge)
    962 		bridge_ifdetach(ifp);
    963 #endif
    964 
    965 	bpf_detach(ifp);
    966 
    967 #if NVLAN > 0
    968 	if (ec->ec_nvlans)
    969 		vlan_ifdetach(ifp);
    970 #endif
    971 
    972 	s = splnet();
    973 	while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) {
    974 		LIST_REMOVE(enm, enm_list);
    975 		free(enm, M_IFMADDR);
    976 		ec->ec_multicnt--;
    977 	}
    978 	splx(s);
    979 
    980 	ifp->if_mowner = NULL;
    981 	MOWNER_DETACH(&ec->ec_rx_mowner);
    982 	MOWNER_DETACH(&ec->ec_tx_mowner);
    983 }
    984 
    985 #if 0
    986 /*
    987  * This is for reference.  We have a table-driven version
    988  * of the little-endian crc32 generator, which is faster
    989  * than the double-loop.
    990  */
    991 uint32_t
    992 ether_crc32_le(const uint8_t *buf, size_t len)
    993 {
    994 	uint32_t c, crc, carry;
    995 	size_t i, j;
    996 
    997 	crc = 0xffffffffU;	/* initial value */
    998 
    999 	for (i = 0; i < len; i++) {
   1000 		c = buf[i];
   1001 		for (j = 0; j < 8; j++) {
   1002 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
   1003 			crc >>= 1;
   1004 			c >>= 1;
   1005 			if (carry)
   1006 				crc = (crc ^ ETHER_CRC_POLY_LE);
   1007 		}
   1008 	}
   1009 
   1010 	return (crc);
   1011 }
   1012 #else
   1013 uint32_t
   1014 ether_crc32_le(const uint8_t *buf, size_t len)
   1015 {
   1016 	static const uint32_t crctab[] = {
   1017 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
   1018 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
   1019 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
   1020 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
   1021 	};
   1022 	uint32_t crc;
   1023 	size_t i;
   1024 
   1025 	crc = 0xffffffffU;	/* initial value */
   1026 
   1027 	for (i = 0; i < len; i++) {
   1028 		crc ^= buf[i];
   1029 		crc = (crc >> 4) ^ crctab[crc & 0xf];
   1030 		crc = (crc >> 4) ^ crctab[crc & 0xf];
   1031 	}
   1032 
   1033 	return (crc);
   1034 }
   1035 #endif
   1036 
   1037 uint32_t
   1038 ether_crc32_be(const uint8_t *buf, size_t len)
   1039 {
   1040 	uint32_t c, crc, carry;
   1041 	size_t i, j;
   1042 
   1043 	crc = 0xffffffffU;	/* initial value */
   1044 
   1045 	for (i = 0; i < len; i++) {
   1046 		c = buf[i];
   1047 		for (j = 0; j < 8; j++) {
   1048 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
   1049 			crc <<= 1;
   1050 			c >>= 1;
   1051 			if (carry)
   1052 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
   1053 		}
   1054 	}
   1055 
   1056 	return (crc);
   1057 }
   1058 
   1059 #ifdef INET
   1060 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] =
   1061     { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 };
   1062 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] =
   1063     { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff };
   1064 #endif
   1065 #ifdef INET6
   1066 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] =
   1067     { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 };
   1068 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] =
   1069     { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff };
   1070 #endif
   1071 
   1072 /*
   1073  * ether_aton implementation, not using a static buffer.
   1074  */
   1075 int
   1076 ether_aton_r(u_char *dest, size_t len, const char *str)
   1077 {
   1078         const u_char *cp = (const void *)str;
   1079 	u_char *ep;
   1080 
   1081 #define atox(c)	(((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10))
   1082 
   1083 	if (len < ETHER_ADDR_LEN)
   1084 		return ENOSPC;
   1085 
   1086 	ep = dest + ETHER_ADDR_LEN;
   1087 
   1088 	while (*cp) {
   1089                 if (!isxdigit(*cp))
   1090                         return EINVAL;
   1091 		*dest = atox(*cp);
   1092 		cp++;
   1093                 if (isxdigit(*cp)) {
   1094                         *dest = (*dest << 4) | atox(*cp);
   1095 			dest++;
   1096 			cp++;
   1097                 } else
   1098 			dest++;
   1099 		if (dest == ep)
   1100 			return *cp == '\0' ? 0 : ENAMETOOLONG;
   1101 		switch (*cp) {
   1102 		case ':':
   1103 		case '-':
   1104 		case '.':
   1105 			cp++;
   1106 			break;
   1107 		}
   1108         }
   1109 	return ENOBUFS;
   1110 }
   1111 
   1112 /*
   1113  * Convert a sockaddr into an Ethernet address or range of Ethernet
   1114  * addresses.
   1115  */
   1116 int
   1117 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN],
   1118     uint8_t addrhi[ETHER_ADDR_LEN])
   1119 {
   1120 #ifdef INET
   1121 	const struct sockaddr_in *sin;
   1122 #endif /* INET */
   1123 #ifdef INET6
   1124 	const struct sockaddr_in6 *sin6;
   1125 #endif /* INET6 */
   1126 
   1127 	switch (sa->sa_family) {
   1128 
   1129 	case AF_UNSPEC:
   1130 		memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN);
   1131 		memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
   1132 		break;
   1133 
   1134 #ifdef INET
   1135 	case AF_INET:
   1136 		sin = satocsin(sa);
   1137 		if (sin->sin_addr.s_addr == INADDR_ANY) {
   1138 			/*
   1139 			 * An IP address of INADDR_ANY means listen to
   1140 			 * or stop listening to all of the Ethernet
   1141 			 * multicast addresses used for IP.
   1142 			 * (This is for the sake of IP multicast routers.)
   1143 			 */
   1144 			memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN);
   1145 			memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN);
   1146 		}
   1147 		else {
   1148 			ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo);
   1149 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
   1150 		}
   1151 		break;
   1152 #endif
   1153 #ifdef INET6
   1154 	case AF_INET6:
   1155 		sin6 = satocsin6(sa);
   1156 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
   1157 			/*
   1158 			 * An IP6 address of 0 means listen to or stop
   1159 			 * listening to all of the Ethernet multicast
   1160 			 * address used for IP6.
   1161 			 * (This is used for multicast routers.)
   1162 			 */
   1163 			memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN);
   1164 			memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN);
   1165 		} else {
   1166 			ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo);
   1167 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
   1168 		}
   1169 		break;
   1170 #endif
   1171 
   1172 	default:
   1173 		return EAFNOSUPPORT;
   1174 	}
   1175 	return 0;
   1176 }
   1177 
   1178 /*
   1179  * Add an Ethernet multicast address or range of addresses to the list for a
   1180  * given interface.
   1181  */
   1182 int
   1183 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec)
   1184 {
   1185 	struct ether_multi *enm;
   1186 	u_char addrlo[ETHER_ADDR_LEN];
   1187 	u_char addrhi[ETHER_ADDR_LEN];
   1188 	int s = splnet(), error;
   1189 
   1190 	error = ether_multiaddr(sa, addrlo, addrhi);
   1191 	if (error != 0) {
   1192 		splx(s);
   1193 		return error;
   1194 	}
   1195 
   1196 	/*
   1197 	 * Verify that we have valid Ethernet multicast addresses.
   1198 	 */
   1199 	if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) {
   1200 		splx(s);
   1201 		return EINVAL;
   1202 	}
   1203 	/*
   1204 	 * See if the address range is already in the list.
   1205 	 */
   1206 	ETHER_LOOKUP_MULTI(addrlo, addrhi, ec, enm);
   1207 	if (enm != NULL) {
   1208 		/*
   1209 		 * Found it; just increment the reference count.
   1210 		 */
   1211 		++enm->enm_refcount;
   1212 		splx(s);
   1213 		return 0;
   1214 	}
   1215 	/*
   1216 	 * New address or range; malloc a new multicast record
   1217 	 * and link it into the interface's multicast list.
   1218 	 */
   1219 	enm = (struct ether_multi *)malloc(sizeof(*enm), M_IFMADDR, M_NOWAIT);
   1220 	if (enm == NULL) {
   1221 		splx(s);
   1222 		return ENOBUFS;
   1223 	}
   1224 	memcpy(enm->enm_addrlo, addrlo, 6);
   1225 	memcpy(enm->enm_addrhi, addrhi, 6);
   1226 	enm->enm_refcount = 1;
   1227 	LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list);
   1228 	ec->ec_multicnt++;
   1229 	splx(s);
   1230 	/*
   1231 	 * Return ENETRESET to inform the driver that the list has changed
   1232 	 * and its reception filter should be adjusted accordingly.
   1233 	 */
   1234 	return ENETRESET;
   1235 }
   1236 
   1237 /*
   1238  * Delete a multicast address record.
   1239  */
   1240 int
   1241 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec)
   1242 {
   1243 	struct ether_multi *enm;
   1244 	u_char addrlo[ETHER_ADDR_LEN];
   1245 	u_char addrhi[ETHER_ADDR_LEN];
   1246 	int s = splnet(), error;
   1247 
   1248 	error = ether_multiaddr(sa, addrlo, addrhi);
   1249 	if (error != 0) {
   1250 		splx(s);
   1251 		return (error);
   1252 	}
   1253 
   1254 	/*
   1255 	 * Look ur the address in our list.
   1256 	 */
   1257 	ETHER_LOOKUP_MULTI(addrlo, addrhi, ec, enm);
   1258 	if (enm == NULL) {
   1259 		splx(s);
   1260 		return (ENXIO);
   1261 	}
   1262 	if (--enm->enm_refcount != 0) {
   1263 		/*
   1264 		 * Still some claims to this record.
   1265 		 */
   1266 		splx(s);
   1267 		return (0);
   1268 	}
   1269 	/*
   1270 	 * No remaining claims to this record; unlink and free it.
   1271 	 */
   1272 	LIST_REMOVE(enm, enm_list);
   1273 	free(enm, M_IFMADDR);
   1274 	ec->ec_multicnt--;
   1275 	splx(s);
   1276 	/*
   1277 	 * Return ENETRESET to inform the driver that the list has changed
   1278 	 * and its reception filter should be adjusted accordingly.
   1279 	 */
   1280 	return (ENETRESET);
   1281 }
   1282 
   1283 void
   1284 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb)
   1285 {
   1286 	ec->ec_ifflags_cb = cb;
   1287 }
   1288 
   1289 /*
   1290  * Common ioctls for Ethernet interfaces.  Note, we must be
   1291  * called at splnet().
   1292  */
   1293 int
   1294 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data)
   1295 {
   1296 	struct ethercom *ec = (void *) ifp;
   1297 	struct eccapreq *eccr;
   1298 	struct ifreq *ifr = (struct ifreq *)data;
   1299 	struct if_laddrreq *iflr = data;
   1300 	const struct sockaddr_dl *sdl;
   1301 	static const uint8_t zero[ETHER_ADDR_LEN];
   1302 	int error;
   1303 
   1304 	switch (cmd) {
   1305 	case SIOCINITIFADDR:
   1306 	    {
   1307 		struct ifaddr *ifa = (struct ifaddr *)data;
   1308 		if (ifa->ifa_addr->sa_family != AF_LINK
   1309 		    && (ifp->if_flags & (IFF_UP|IFF_RUNNING)) !=
   1310 		       (IFF_UP|IFF_RUNNING)) {
   1311 			ifp->if_flags |= IFF_UP;
   1312 			if ((error = (*ifp->if_init)(ifp)) != 0)
   1313 				return error;
   1314 		}
   1315 #ifdef INET
   1316 		if (ifa->ifa_addr->sa_family == AF_INET)
   1317 			arp_ifinit(ifp, ifa);
   1318 #endif /* INET */
   1319 		return 0;
   1320 	    }
   1321 
   1322 	case SIOCSIFMTU:
   1323 	    {
   1324 		int maxmtu;
   1325 
   1326 		if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU)
   1327 			maxmtu = ETHERMTU_JUMBO;
   1328 		else
   1329 			maxmtu = ETHERMTU;
   1330 
   1331 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu)
   1332 			return EINVAL;
   1333 		else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET)
   1334 			return error;
   1335 		else if (ifp->if_flags & IFF_UP) {
   1336 			/* Make sure the device notices the MTU change. */
   1337 			return (*ifp->if_init)(ifp);
   1338 		} else
   1339 			return 0;
   1340 	    }
   1341 
   1342 	case SIOCSIFFLAGS:
   1343 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
   1344 			return error;
   1345 		switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
   1346 		case IFF_RUNNING:
   1347 			/*
   1348 			 * If interface is marked down and it is running,
   1349 			 * then stop and disable it.
   1350 			 */
   1351 			(*ifp->if_stop)(ifp, 1);
   1352 			break;
   1353 		case IFF_UP:
   1354 			/*
   1355 			 * If interface is marked up and it is stopped, then
   1356 			 * start it.
   1357 			 */
   1358 			return (*ifp->if_init)(ifp);
   1359 		case IFF_UP|IFF_RUNNING:
   1360 			error = 0;
   1361 			if (ec->ec_ifflags_cb == NULL ||
   1362 			    (error = (*ec->ec_ifflags_cb)(ec)) == ENETRESET) {
   1363 				/*
   1364 				 * Reset the interface to pick up
   1365 				 * changes in any other flags that
   1366 				 * affect the hardware state.
   1367 				 */
   1368 				return (*ifp->if_init)(ifp);
   1369 			} else
   1370 				return error;
   1371 		case 0:
   1372 			break;
   1373 		}
   1374 		return 0;
   1375 	case SIOCGETHERCAP:
   1376 		eccr = (struct eccapreq *)data;
   1377 		eccr->eccr_capabilities = ec->ec_capabilities;
   1378 		eccr->eccr_capenable = ec->ec_capenable;
   1379 		return 0;
   1380 	case SIOCADDMULTI:
   1381 		return ether_addmulti(ifreq_getaddr(cmd, ifr), ec);
   1382 	case SIOCDELMULTI:
   1383 		return ether_delmulti(ifreq_getaddr(cmd, ifr), ec);
   1384 	case SIOCSIFMEDIA:
   1385 	case SIOCGIFMEDIA:
   1386 		if (ec->ec_mii == NULL)
   1387 			return ENOTTY;
   1388 		return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media, cmd);
   1389 	case SIOCALIFADDR:
   1390 		sdl = satocsdl(sstocsa(&iflr->addr));
   1391 		if (sdl->sdl_family != AF_LINK)
   1392 			;
   1393 		else if (ETHER_IS_MULTICAST(CLLADDR(sdl)))
   1394 			return EINVAL;
   1395 		else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0)
   1396 			return EINVAL;
   1397 		/*FALLTHROUGH*/
   1398 	default:
   1399 		return ifioctl_common(ifp, cmd, data);
   1400 	}
   1401 	return 0;
   1402 }
   1403 
   1404 /*
   1405  * Enable/disable passing VLAN packets if the parent interface supports it.
   1406  * Return:
   1407  * 	 0: Ok
   1408  *	-1: Parent interface does not support vlans
   1409  *	>0: Error
   1410  */
   1411 int
   1412 ether_enable_vlan_mtu(struct ifnet *ifp)
   1413 {
   1414 	int error;
   1415 	struct ethercom *ec = (void *)ifp;
   1416 
   1417 	/* Already have VLAN's do nothing. */
   1418 	if (ec->ec_nvlans != 0)
   1419 		return 0;
   1420 
   1421 	/* Parent does not support VLAN's */
   1422 	if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0)
   1423 		return -1;
   1424 
   1425 	/*
   1426 	 * Parent supports the VLAN_MTU capability,
   1427 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames;
   1428 	 * enable it.
   1429 	 */
   1430 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
   1431 
   1432 	/* Interface is down, defer for later */
   1433 	if ((ifp->if_flags & IFF_UP) == 0)
   1434 		return 0;
   1435 
   1436 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
   1437 		return 0;
   1438 
   1439 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
   1440 	return error;
   1441 }
   1442 
   1443 int
   1444 ether_disable_vlan_mtu(struct ifnet *ifp)
   1445 {
   1446 	int error;
   1447 	struct ethercom *ec = (void *)ifp;
   1448 
   1449 	/* We still have VLAN's, defer for later */
   1450 	if (ec->ec_nvlans != 0)
   1451 		return 0;
   1452 
   1453 	/* Parent does not support VLAB's, nothing to do. */
   1454 	if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0)
   1455 		return -1;
   1456 
   1457 	/*
   1458 	 * Disable Tx/Rx of VLAN-sized frames.
   1459 	 */
   1460 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
   1461 
   1462 	/* Interface is down, defer for later */
   1463 	if ((ifp->if_flags & IFF_UP) == 0)
   1464 		return 0;
   1465 
   1466 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
   1467 		return 0;
   1468 
   1469 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
   1470 	return error;
   1471 }
   1472 
   1473 static int
   1474 ether_multicast_sysctl(SYSCTLFN_ARGS)
   1475 {
   1476 	struct ether_multi *enm;
   1477 	struct ether_multi_sysctl addr;
   1478 	struct ifnet *ifp;
   1479 	struct ethercom *ec;
   1480 	int error;
   1481 	size_t written;
   1482 
   1483 	if (namelen != 1)
   1484 		return EINVAL;
   1485 
   1486 	ifp = if_byindex(name[0]);
   1487 	if (ifp == NULL)
   1488 		return ENODEV;
   1489 	if (ifp->if_type != IFT_ETHER) {
   1490 		*oldlenp = 0;
   1491 		return 0;
   1492 	}
   1493 	ec = (struct ethercom *)ifp;
   1494 
   1495 	if (oldp == NULL) {
   1496 		*oldlenp = ec->ec_multicnt * sizeof(addr);
   1497 		return 0;
   1498 	}
   1499 
   1500 	memset(&addr, 0, sizeof(addr));
   1501 	error = 0;
   1502 	written = 0;
   1503 
   1504 	LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) {
   1505 		if (written + sizeof(addr) > *oldlenp)
   1506 			break;
   1507 		addr.enm_refcount = enm->enm_refcount;
   1508 		memcpy(addr.enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN);
   1509 		memcpy(addr.enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN);
   1510 		error = sysctl_copyout(l, &addr, oldp, sizeof(addr));
   1511 		if (error)
   1512 			break;
   1513 		written += sizeof(addr);
   1514 		oldp = (char *)oldp + sizeof(addr);
   1515 	}
   1516 
   1517 	*oldlenp = written;
   1518 	return error;
   1519 }
   1520 
   1521 SYSCTL_SETUP(sysctl_net_ether_setup, "sysctl net.ether subtree setup")
   1522 {
   1523 	const struct sysctlnode *rnode = NULL;
   1524 
   1525 	sysctl_createv(clog, 0, NULL, &rnode,
   1526 		       CTLFLAG_PERMANENT,
   1527 		       CTLTYPE_NODE, "ether",
   1528 		       SYSCTL_DESCR("Ethernet-specific information"),
   1529 		       NULL, 0, NULL, 0,
   1530 		       CTL_NET, CTL_CREATE, CTL_EOL);
   1531 
   1532 	sysctl_createv(clog, 0, &rnode, NULL,
   1533 		       CTLFLAG_PERMANENT,
   1534 		       CTLTYPE_NODE, "multicast",
   1535 		       SYSCTL_DESCR("multicast addresses"),
   1536 		       ether_multicast_sysctl, 0, NULL, 0,
   1537 		       CTL_CREATE, CTL_EOL);
   1538 }
   1539 
   1540 void
   1541 etherinit(void)
   1542 {
   1543 	mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET);
   1544 }
   1545