Home | History | Annotate | Line # | Download | only in net
if_ethersubr.c revision 1.224
      1 /*	$NetBSD: if_ethersubr.c,v 1.224 2016/06/20 07:01:45 knakahara 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 1.224 2016/06/20 07:01:45 knakahara 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 *, 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 	KASSERT(ifp->if_extflags & IFEF_OUTPUT_MPSAFE);
    208 
    209 #ifdef MBUFTRACE
    210 	m_claimm(m, ifp->if_mowner);
    211 #endif
    212 
    213 #if NCARP > 0
    214 	if (ifp->if_type == IFT_CARP) {
    215 		struct ifaddr *ifa;
    216 
    217 		/* loop back if this is going to the carp interface */
    218 		if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP &&
    219 		    (ifa = ifa_ifwithaddr(dst)) != NULL &&
    220 		    ifa->ifa_ifp == ifp0)
    221 			return looutput(ifp0, m, dst, rt);
    222 
    223 		ifp = ifp->if_carpdev;
    224 		/* ac = (struct arpcom *)ifp; */
    225 
    226 		if ((ifp0->if_flags & (IFF_UP|IFF_RUNNING)) !=
    227 		    (IFF_UP|IFF_RUNNING))
    228 			senderr(ENETDOWN);
    229 	}
    230 #endif /* NCARP > 0 */
    231 
    232 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
    233 		senderr(ENETDOWN);
    234 
    235 	switch (dst->sa_family) {
    236 
    237 #ifdef INET
    238 	case AF_INET:
    239 		KERNEL_LOCK(1, NULL);
    240 		if (m->m_flags & M_BCAST)
    241 			(void)memcpy(edst, etherbroadcastaddr, sizeof(edst));
    242 		else if (m->m_flags & M_MCAST)
    243 			ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst);
    244 		else if ((error = arpresolve(ifp, rt, m, dst, edst,
    245 		    sizeof(edst))) != 0) {
    246 			KERNEL_UNLOCK_ONE(NULL);
    247 			return error == EWOULDBLOCK ? 0 : error;
    248 		}
    249 		KERNEL_UNLOCK_ONE(NULL);
    250 		/* If broadcasting on a simplex interface, loopback a copy */
    251 		if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
    252 			mcopy = m_copy(m, 0, (int)M_COPYALL);
    253 		etype = htons(ETHERTYPE_IP);
    254 		break;
    255 
    256 	case AF_ARP:
    257 		ah = mtod(m, struct arphdr *);
    258 		if (m->m_flags & M_BCAST)
    259 			(void)memcpy(edst, etherbroadcastaddr, sizeof(edst));
    260 		else {
    261 			void *tha = ar_tha(ah);
    262 
    263 			if (tha == NULL) {
    264 				/* fake with ARPHDR_IEEE1394 */
    265 				return 0;
    266 			}
    267 			memcpy(edst, tha, sizeof(edst));
    268 		}
    269 
    270 		ah->ar_hrd = htons(ARPHRD_ETHER);
    271 
    272 		switch (ntohs(ah->ar_op)) {
    273 		case ARPOP_REVREQUEST:
    274 		case ARPOP_REVREPLY:
    275 			etype = htons(ETHERTYPE_REVARP);
    276 			break;
    277 
    278 		case ARPOP_REQUEST:
    279 		case ARPOP_REPLY:
    280 		default:
    281 			etype = htons(ETHERTYPE_ARP);
    282 		}
    283 
    284 		break;
    285 #endif
    286 #ifdef INET6
    287 	case AF_INET6:
    288 		if (!nd6_storelladdr(ifp, rt, m, dst, edst, sizeof(edst))){
    289 			/* something bad happened */
    290 			return (0);
    291 		}
    292 		etype = htons(ETHERTYPE_IPV6);
    293 		break;
    294 #endif
    295 #ifdef NETATALK
    296     case AF_APPLETALK:
    297 		KERNEL_LOCK(1, NULL);
    298 		if (aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) {
    299 #ifdef NETATALKDEBUG
    300 			printf("aarpresolv failed\n");
    301 #endif /* NETATALKDEBUG */
    302 			KERNEL_UNLOCK_ONE(NULL);
    303 			return (0);
    304 		}
    305 		/*
    306 		 * ifaddr is the first thing in at_ifaddr
    307 		 */
    308 		aa = (struct at_ifaddr *) at_ifawithnet(
    309 		    (const struct sockaddr_at *)dst, ifp);
    310 		if (aa == NULL) {
    311 		    KERNEL_UNLOCK_ONE(NULL);
    312 		    goto bad;
    313 		}
    314 
    315 		/*
    316 		 * In the phase 2 case, we need to prepend an mbuf for the
    317 		 * llc header.  Since we must preserve the value of m,
    318 		 * which is passed to us by value, we m_copy() the first
    319 		 * mbuf, and use it for our llc header.
    320 		 */
    321 		if (aa->aa_flags & AFA_PHASE2) {
    322 			struct llc llc;
    323 
    324 			M_PREPEND(m, sizeof(struct llc), M_DONTWAIT);
    325 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
    326 			llc.llc_control = LLC_UI;
    327 			memcpy(llc.llc_snap_org_code, at_org_code,
    328 			    sizeof(llc.llc_snap_org_code));
    329 			llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK);
    330 			memcpy(mtod(m, void *), &llc, sizeof(struct llc));
    331 		} else {
    332 			etype = htons(ETHERTYPE_ATALK);
    333 		}
    334 		KERNEL_UNLOCK_ONE(NULL);
    335 		break;
    336 #endif /* NETATALK */
    337 	case pseudo_AF_HDRCMPLT:
    338 		hdrcmplt = 1;
    339 		memcpy(esrc,
    340 		    ((const struct ether_header *)dst->sa_data)->ether_shost,
    341 		    sizeof(esrc));
    342 		/* FALLTHROUGH */
    343 
    344 	case AF_UNSPEC:
    345  		memcpy(edst,
    346 		    ((const struct ether_header *)dst->sa_data)->ether_dhost,
    347 		    sizeof(edst));
    348 		/* AF_UNSPEC doesn't swap the byte order of the ether_type. */
    349 		etype = ((const struct ether_header *)dst->sa_data)->ether_type;
    350 		break;
    351 
    352 	default:
    353 		printf("%s: can't handle af%d\n", ifp->if_xname,
    354 			dst->sa_family);
    355 		senderr(EAFNOSUPPORT);
    356 	}
    357 
    358 #ifdef MPLS
    359 	KERNEL_LOCK(1, NULL);
    360 	{
    361 		struct m_tag *mtag;
    362 		mtag = m_tag_find(m, PACKET_TAG_MPLS, NULL);
    363 		if (mtag != NULL) {
    364 			/* Having the tag itself indicates it's MPLS */
    365 			etype = htons(ETHERTYPE_MPLS);
    366 			m_tag_delete(m, mtag);
    367 		}
    368 	}
    369 	KERNEL_UNLOCK_ONE(NULL);
    370 #endif
    371 
    372 	if (mcopy)
    373 		(void)looutput(ifp, mcopy, dst, rt);
    374 
    375 	/* If no ether type is set, this must be a 802.2 formatted packet.
    376 	 */
    377 	if (etype == 0)
    378 		etype = htons(m->m_pkthdr.len);
    379 	/*
    380 	 * Add local net header.  If no space in first mbuf,
    381 	 * allocate another.
    382 	 */
    383 	M_PREPEND(m, sizeof (struct ether_header), M_DONTWAIT);
    384 	if (m == 0)
    385 		senderr(ENOBUFS);
    386 	eh = mtod(m, struct ether_header *);
    387 	/* Note: etype is already in network byte order. */
    388 	(void)memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
    389  	memcpy(eh->ether_dhost, edst, sizeof(edst));
    390 	if (hdrcmplt)
    391 		memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost));
    392 	else
    393 	 	memcpy(eh->ether_shost, CLLADDR(ifp->if_sadl),
    394 		    sizeof(eh->ether_shost));
    395 
    396 #if NCARP > 0
    397 	if (ifp0 != ifp && ifp0->if_type == IFT_CARP) {
    398 	 	memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl),
    399 		    sizeof(eh->ether_shost));
    400 	}
    401 #endif /* NCARP > 0 */
    402 
    403 	if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0)
    404 		return (error);
    405 	if (m == NULL)
    406 		return (0);
    407 
    408 #if NBRIDGE > 0
    409 	/*
    410 	 * Bridges require special output handling.
    411 	 */
    412 	if (ifp->if_bridge)
    413 		return (bridge_output(ifp, m, NULL, NULL));
    414 #endif
    415 
    416 #if NCARP > 0
    417 	if (ifp != ifp0)
    418 		ifp0->if_obytes += m->m_pkthdr.len + ETHER_HDR_LEN;
    419 #endif /* NCARP > 0 */
    420 
    421 #ifdef ALTQ
    422 	KERNEL_LOCK(1, NULL);
    423 	/*
    424 	 * If ALTQ is enabled on the parent interface, do
    425 	 * classification; the queueing discipline might not
    426 	 * require classification, but might require the
    427 	 * address family/header pointer in the pktattr.
    428 	 */
    429 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
    430 		altq_etherclassify(&ifp->if_snd, m);
    431 	KERNEL_UNLOCK_ONE(NULL);
    432 #endif
    433 	return ifq_enqueue(ifp, m);
    434 
    435 bad:
    436 	if (m)
    437 		m_freem(m);
    438 	return (error);
    439 }
    440 
    441 #ifdef ALTQ
    442 /*
    443  * This routine is a slight hack to allow a packet to be classified
    444  * if the Ethernet headers are present.  It will go away when ALTQ's
    445  * classification engine understands link headers.
    446  */
    447 void
    448 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m)
    449 {
    450 	struct ether_header *eh;
    451 	uint16_t ether_type;
    452 	int hlen, af, hdrsize;
    453 	void *hdr;
    454 
    455 	hlen = ETHER_HDR_LEN;
    456 	eh = mtod(m, struct ether_header *);
    457 
    458 	ether_type = htons(eh->ether_type);
    459 
    460 	if (ether_type < ETHERMTU) {
    461 		/* LLC/SNAP */
    462 		struct llc *llc = (struct llc *)(eh + 1);
    463 		hlen += 8;
    464 
    465 		if (m->m_len < hlen ||
    466 		    llc->llc_dsap != LLC_SNAP_LSAP ||
    467 		    llc->llc_ssap != LLC_SNAP_LSAP ||
    468 		    llc->llc_control != LLC_UI) {
    469 			/* Not SNAP. */
    470 			goto bad;
    471 		}
    472 
    473 		ether_type = htons(llc->llc_un.type_snap.ether_type);
    474 	}
    475 
    476 	switch (ether_type) {
    477 	case ETHERTYPE_IP:
    478 		af = AF_INET;
    479 		hdrsize = 20;		/* sizeof(struct ip) */
    480 		break;
    481 
    482 	case ETHERTYPE_IPV6:
    483 		af = AF_INET6;
    484 		hdrsize = 40;		/* sizeof(struct ip6_hdr) */
    485 		break;
    486 
    487 	default:
    488 		af = AF_UNSPEC;
    489 		hdrsize = 0;
    490 		break;
    491 	}
    492 
    493 	while (m->m_len <= hlen) {
    494 		hlen -= m->m_len;
    495 		m = m->m_next;
    496 	}
    497 	if (m->m_len < (hlen + hdrsize)) {
    498 		/*
    499 		 * protocol header not in a single mbuf.
    500 		 * We can't cope with this situation right
    501 		 * now (but it shouldn't ever happen, really, anyhow).
    502 		 */
    503 #ifdef DEBUG
    504 		printf("altq_etherclassify: headers span multiple mbufs: "
    505 		    "%d < %d\n", m->m_len, (hlen + hdrsize));
    506 #endif
    507 		goto bad;
    508 	}
    509 
    510 	m->m_data += hlen;
    511 	m->m_len -= hlen;
    512 
    513 	hdr = mtod(m, void *);
    514 
    515 	if (ALTQ_NEEDS_CLASSIFY(ifq))
    516 		m->m_pkthdr.pattr_class =
    517 		    (*ifq->altq_classify)(ifq->altq_clfier, m, af);
    518 	m->m_pkthdr.pattr_af = af;
    519 	m->m_pkthdr.pattr_hdr = hdr;
    520 
    521 	m->m_data -= hlen;
    522 	m->m_len += hlen;
    523 
    524 	return;
    525 
    526  bad:
    527 	m->m_pkthdr.pattr_class = NULL;
    528 	m->m_pkthdr.pattr_hdr = NULL;
    529 	m->m_pkthdr.pattr_af = AF_UNSPEC;
    530 }
    531 #endif /* ALTQ */
    532 
    533 /*
    534  * Process a received Ethernet packet;
    535  * the packet is in the mbuf chain m with
    536  * the ether header.
    537  */
    538 void
    539 ether_input(struct ifnet *ifp, struct mbuf *m)
    540 {
    541 	struct ethercom *ec = (struct ethercom *) ifp;
    542 	pktqueue_t *pktq = NULL;
    543 	struct ifqueue *inq = NULL;
    544 	uint16_t etype;
    545 	struct ether_header *eh;
    546 	size_t ehlen;
    547 	static int earlypkts;
    548 	int isr = 0;
    549 #if defined (LLC) || defined(NETATALK)
    550 	struct llc *l;
    551 #endif
    552 
    553 	KASSERT(!cpu_intr_p());
    554 
    555 	if ((ifp->if_flags & IFF_UP) == 0) {
    556 		m_freem(m);
    557 		return;
    558 	}
    559 
    560 #ifdef MBUFTRACE
    561 	m_claimm(m, &ec->ec_rx_mowner);
    562 #endif
    563 	eh = mtod(m, struct ether_header *);
    564 	etype = ntohs(eh->ether_type);
    565 	ehlen = sizeof(*eh);
    566 
    567 	if(__predict_false(earlypkts < 100 || !rnd_initial_entropy)) {
    568 		rnd_add_data(NULL, eh, ehlen, 0);
    569 		earlypkts++;
    570 	}
    571 
    572 	/*
    573 	 * Determine if the packet is within its size limits.
    574 	 */
    575 	if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len >
    576 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
    577 		mutex_enter(&bigpktpps_lock);
    578 		if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count,
    579 			    bigpktppslim)) {
    580 			printf("%s: discarding oversize frame (len=%d)\n",
    581 			    ifp->if_xname, m->m_pkthdr.len);
    582 		}
    583 		mutex_exit(&bigpktpps_lock);
    584 		m_freem(m);
    585 		return;
    586 	}
    587 
    588 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
    589 		/*
    590 		 * If this is not a simplex interface, drop the packet
    591 		 * if it came from us.
    592 		 */
    593 		if ((ifp->if_flags & IFF_SIMPLEX) == 0 &&
    594 		    memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost,
    595 		    ETHER_ADDR_LEN) == 0) {
    596 			m_freem(m);
    597 			return;
    598 		}
    599 
    600 		if (memcmp(etherbroadcastaddr,
    601 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
    602 			m->m_flags |= M_BCAST;
    603 		else
    604 			m->m_flags |= M_MCAST;
    605 		ifp->if_imcasts++;
    606 	}
    607 
    608 	/* If the CRC is still on the packet, trim it off. */
    609 	if (m->m_flags & M_HASFCS) {
    610 		m_adj(m, -ETHER_CRC_LEN);
    611 		m->m_flags &= ~M_HASFCS;
    612 	}
    613 
    614 	ifp->if_ibytes += m->m_pkthdr.len;
    615 
    616 #if NCARP > 0
    617 	if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) {
    618 		/*
    619 		 * clear M_PROMISC, in case the packets comes from a
    620 		 * vlan
    621 		 */
    622 		m->m_flags &= ~M_PROMISC;
    623 		if (carp_input(m, (uint8_t *)&eh->ether_shost,
    624 		    (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0)
    625 			return;
    626 	}
    627 #endif /* NCARP > 0 */
    628 	if ((m->m_flags & (M_BCAST|M_MCAST|M_PROMISC)) == 0 &&
    629 	    (ifp->if_flags & IFF_PROMISC) != 0 &&
    630 	    memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost,
    631 		   ETHER_ADDR_LEN) != 0) {
    632 		m->m_flags |= M_PROMISC;
    633 	}
    634 
    635 	if ((m->m_flags & M_PROMISC) == 0) {
    636 		if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
    637 			return;
    638 		if (m == NULL)
    639 			return;
    640 
    641 		eh = mtod(m, struct ether_header *);
    642 		etype = ntohs(eh->ether_type);
    643 		ehlen = sizeof(*eh);
    644 	}
    645 
    646 #if NAGR > 0
    647 	if (ifp->if_agrprivate &&
    648 	    __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) {
    649 		m->m_flags &= ~M_PROMISC;
    650 		agr_input(ifp, m);
    651 		return;
    652 	}
    653 #endif /* NAGR > 0 */
    654 
    655 	/*
    656 	 * If VLANs are configured on the interface, check to
    657 	 * see if the device performed the decapsulation and
    658 	 * provided us with the tag.
    659 	 */
    660 	if (ec->ec_nvlans && m_tag_find(m, PACKET_TAG_VLAN, NULL) != NULL) {
    661 #if NVLAN > 0
    662 		/*
    663 		 * vlan_input() will either recursively call ether_input()
    664 		 * or drop the packet.
    665 		 */
    666 		vlan_input(ifp, m);
    667 #else
    668 		m_freem(m);
    669 #endif
    670 		return;
    671 	}
    672 
    673 	/*
    674 	 * Handle protocols that expect to have the Ethernet header
    675 	 * (and possibly FCS) intact.
    676 	 */
    677 	switch (etype) {
    678 	case ETHERTYPE_VLAN: {
    679 		struct ether_vlan_header *evl = (void *)eh;
    680 		/*
    681 		 * If there is a tag of 0, then the VLAN header was probably
    682 		 * just being used to store the priority.  Extract the ether
    683 		 * type, and if IP or IPV6, let them deal with it.
    684 		 */
    685 		if (m->m_len <= sizeof(*evl)
    686 		    && EVL_VLANOFTAG(evl->evl_tag) == 0) {
    687 			etype = ntohs(evl->evl_proto);
    688 			ehlen = sizeof(*evl);
    689 			if ((m->m_flags & M_PROMISC) == 0
    690 			    && (etype == ETHERTYPE_IP
    691 				|| etype == ETHERTYPE_IPV6))
    692 				break;
    693 		}
    694 #if NVLAN > 0
    695 		/*
    696 		 * vlan_input() will either recursively call ether_input()
    697 		 * or drop the packet.
    698 		 */
    699 		if (((struct ethercom *)ifp)->ec_nvlans != 0)
    700 			vlan_input(ifp, m);
    701 		else
    702 #endif /* NVLAN > 0 */
    703 			m_freem(m);
    704 		return;
    705 	}
    706 #if NPPPOE > 0
    707 	case ETHERTYPE_PPPOEDISC:
    708 		pppoedisc_input(ifp, m);
    709 		return;
    710 	case ETHERTYPE_PPPOE:
    711 		pppoe_input(ifp, m);
    712 		return;
    713 #endif /* NPPPOE > 0 */
    714 	case ETHERTYPE_SLOWPROTOCOLS: {
    715 		uint8_t subtype;
    716 
    717 #if defined(DIAGNOSTIC)
    718 		if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype)) {
    719 			panic("ether_input: too short slow protocol packet");
    720 		}
    721 #endif
    722 		m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype);
    723 		switch (subtype) {
    724 #if NAGR > 0
    725 		case SLOWPROTOCOLS_SUBTYPE_LACP:
    726 			if (ifp->if_agrprivate) {
    727 				ieee8023ad_lacp_input(ifp, m);
    728 				return;
    729 			}
    730 			break;
    731 
    732 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
    733 			if (ifp->if_agrprivate) {
    734 				ieee8023ad_marker_input(ifp, m);
    735 				return;
    736 			}
    737 			break;
    738 #endif /* NAGR > 0 */
    739 		default:
    740 			if (subtype == 0 || subtype > 10) {
    741 				/* illegal value */
    742 				m_freem(m);
    743 				return;
    744 			}
    745 			/* unknown subtype */
    746 			break;
    747 		}
    748 		/* FALLTHROUGH */
    749 	}
    750 	default:
    751 		if (m->m_flags & M_PROMISC) {
    752 			m_freem(m);
    753 			return;
    754 		}
    755 	}
    756 
    757 	/* If the CRC is still on the packet, trim it off. */
    758 	if (m->m_flags & M_HASFCS) {
    759 		m_adj(m, -ETHER_CRC_LEN);
    760 		m->m_flags &= ~M_HASFCS;
    761 	}
    762 
    763 	if (etype > ETHERMTU + sizeof (struct ether_header)) {
    764 		/* Strip off the Ethernet header. */
    765 		m_adj(m, ehlen);
    766 
    767 		switch (etype) {
    768 #ifdef INET
    769 		case ETHERTYPE_IP:
    770 #ifdef GATEWAY
    771 			if (ipflow_fastforward(m))
    772 				return;
    773 #endif
    774 			pktq = ip_pktq;
    775 			break;
    776 
    777 		case ETHERTYPE_ARP:
    778 			isr = NETISR_ARP;
    779 			inq = &arpintrq;
    780 			break;
    781 
    782 		case ETHERTYPE_REVARP:
    783 			revarpinput(m);	/* XXX queue? */
    784 			return;
    785 #endif
    786 #ifdef INET6
    787 		case ETHERTYPE_IPV6:
    788 			if (__predict_false(!in6_present)) {
    789 				m_freem(m);
    790 				return;
    791 			}
    792 #ifdef GATEWAY
    793 			if (ip6flow_fastforward(&m))
    794 				return;
    795 #endif
    796 			pktq = ip6_pktq;
    797 			break;
    798 #endif
    799 #ifdef NETATALK
    800 		case ETHERTYPE_ATALK:
    801 			isr = NETISR_ATALK;
    802 			inq = &atintrq1;
    803 			break;
    804 		case ETHERTYPE_AARP:
    805 			/* probably this should be done with a NETISR as well */
    806 			aarpinput(ifp, m); /* XXX */
    807 			return;
    808 #endif /* NETATALK */
    809 #ifdef MPLS
    810 		case ETHERTYPE_MPLS:
    811 			isr = NETISR_MPLS;
    812 			inq = &mplsintrq;
    813 			break;
    814 #endif
    815 		default:
    816 			m_freem(m);
    817 			return;
    818 		}
    819 	} else {
    820 #if defined (LLC) || defined (NETATALK)
    821 		l = (struct llc *)(eh+1);
    822 		switch (l->llc_dsap) {
    823 #ifdef NETATALK
    824 		case LLC_SNAP_LSAP:
    825 			switch (l->llc_control) {
    826 			case LLC_UI:
    827 				if (l->llc_ssap != LLC_SNAP_LSAP) {
    828 					goto dropanyway;
    829 				}
    830 
    831 				if (memcmp(&(l->llc_snap_org_code)[0],
    832 				    at_org_code, sizeof(at_org_code)) == 0 &&
    833 				    ntohs(l->llc_snap_ether_type) ==
    834 				    ETHERTYPE_ATALK) {
    835 					inq = &atintrq2;
    836 					m_adj(m, sizeof(struct ether_header)
    837 					    + sizeof(struct llc));
    838 					isr = NETISR_ATALK;
    839 					break;
    840 				}
    841 
    842 				if (memcmp(&(l->llc_snap_org_code)[0],
    843 				    aarp_org_code,
    844 				    sizeof(aarp_org_code)) == 0 &&
    845 				    ntohs(l->llc_snap_ether_type) ==
    846 				    ETHERTYPE_AARP) {
    847 					m_adj( m, sizeof(struct ether_header)
    848 					    + sizeof(struct llc));
    849 					aarpinput(ifp, m); /* XXX */
    850 				    return;
    851 				}
    852 
    853 			default:
    854 				goto dropanyway;
    855 			}
    856 			break;
    857 		dropanyway:
    858 #endif
    859 		default:
    860 			m_freem(m);
    861 			return;
    862 		}
    863 #else /* ISO || LLC || NETATALK*/
    864 		m_freem(m);
    865 		return;
    866 #endif /* ISO || LLC || NETATALK*/
    867 	}
    868 
    869 	if (__predict_true(pktq)) {
    870 		const uint32_t h = pktq_rps_hash(m);
    871 		if (__predict_false(!pktq_enqueue(pktq, m, h))) {
    872 			m_freem(m);
    873 		}
    874 		return;
    875 	}
    876 
    877 	if (__predict_false(!inq)) {
    878 		/* Should not happen. */
    879 		m_freem(m);
    880 		return;
    881 	}
    882 	if (IF_QFULL(inq)) {
    883 		IF_DROP(inq);
    884 		m_freem(m);
    885 	} else {
    886 		IF_ENQUEUE(inq, m);
    887 		schednetisr(isr);
    888 	}
    889 }
    890 
    891 /*
    892  * Convert Ethernet address to printable (loggable) representation.
    893  */
    894 char *
    895 ether_sprintf(const u_char *ap)
    896 {
    897 	static char etherbuf[3 * ETHER_ADDR_LEN];
    898 	return ether_snprintf(etherbuf, sizeof(etherbuf), ap);
    899 }
    900 
    901 char *
    902 ether_snprintf(char *buf, size_t len, const u_char *ap)
    903 {
    904 	char *cp = buf;
    905 	size_t i;
    906 
    907 	for (i = 0; i < len / 3; i++) {
    908 		*cp++ = hexdigits[*ap >> 4];
    909 		*cp++ = hexdigits[*ap++ & 0xf];
    910 		*cp++ = ':';
    911 	}
    912 	*--cp = '\0';
    913 	return buf;
    914 }
    915 
    916 /*
    917  * Perform common duties while attaching to interface list
    918  */
    919 void
    920 ether_ifattach(struct ifnet *ifp, const uint8_t *lla)
    921 {
    922 	struct ethercom *ec = (struct ethercom *)ifp;
    923 
    924 	ifp->if_extflags |= IFEF_OUTPUT_MPSAFE;
    925 	ifp->if_type = IFT_ETHER;
    926 	ifp->if_hdrlen = ETHER_HDR_LEN;
    927 	ifp->if_dlt = DLT_EN10MB;
    928 	ifp->if_mtu = ETHERMTU;
    929 	ifp->if_output = ether_output;
    930 	ifp->_if_input = ether_input;
    931 	if (ifp->if_baudrate == 0)
    932 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
    933 
    934 	if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla));
    935 
    936 	LIST_INIT(&ec->ec_multiaddrs);
    937 	ifp->if_broadcastaddr = etherbroadcastaddr;
    938 	bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header));
    939 #ifdef MBUFTRACE
    940 	strlcpy(ec->ec_tx_mowner.mo_name, ifp->if_xname,
    941 	    sizeof(ec->ec_tx_mowner.mo_name));
    942 	strlcpy(ec->ec_tx_mowner.mo_descr, "tx",
    943 	    sizeof(ec->ec_tx_mowner.mo_descr));
    944 	strlcpy(ec->ec_rx_mowner.mo_name, ifp->if_xname,
    945 	    sizeof(ec->ec_rx_mowner.mo_name));
    946 	strlcpy(ec->ec_rx_mowner.mo_descr, "rx",
    947 	    sizeof(ec->ec_rx_mowner.mo_descr));
    948 	MOWNER_ATTACH(&ec->ec_tx_mowner);
    949 	MOWNER_ATTACH(&ec->ec_rx_mowner);
    950 	ifp->if_mowner = &ec->ec_tx_mowner;
    951 #endif
    952 }
    953 
    954 void
    955 ether_ifdetach(struct ifnet *ifp)
    956 {
    957 	struct ethercom *ec = (void *) ifp;
    958 	struct ether_multi *enm;
    959 	int s;
    960 
    961 	/*
    962 	 * Prevent further calls to ioctl (for example turning off
    963 	 * promiscuous mode from the bridge code), which eventually can
    964 	 * call if_init() which can cause panics because the interface
    965 	 * is in the process of being detached. Return device not configured
    966 	 * instead.
    967 	 */
    968 	ifp->if_ioctl = (int (*)(struct ifnet *, u_long, void *))enxio;
    969 
    970 #if NBRIDGE > 0
    971 	if (ifp->if_bridge)
    972 		bridge_ifdetach(ifp);
    973 #endif
    974 
    975 	bpf_detach(ifp);
    976 
    977 #if NVLAN > 0
    978 	if (ec->ec_nvlans)
    979 		vlan_ifdetach(ifp);
    980 #endif
    981 
    982 	s = splnet();
    983 	while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) {
    984 		LIST_REMOVE(enm, enm_list);
    985 		free(enm, M_IFMADDR);
    986 		ec->ec_multicnt--;
    987 	}
    988 	splx(s);
    989 
    990 	ifp->if_mowner = NULL;
    991 	MOWNER_DETACH(&ec->ec_rx_mowner);
    992 	MOWNER_DETACH(&ec->ec_tx_mowner);
    993 }
    994 
    995 #if 0
    996 /*
    997  * This is for reference.  We have a table-driven version
    998  * of the little-endian crc32 generator, which is faster
    999  * than the double-loop.
   1000  */
   1001 uint32_t
   1002 ether_crc32_le(const uint8_t *buf, size_t len)
   1003 {
   1004 	uint32_t c, crc, carry;
   1005 	size_t i, j;
   1006 
   1007 	crc = 0xffffffffU;	/* initial value */
   1008 
   1009 	for (i = 0; i < len; i++) {
   1010 		c = buf[i];
   1011 		for (j = 0; j < 8; j++) {
   1012 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
   1013 			crc >>= 1;
   1014 			c >>= 1;
   1015 			if (carry)
   1016 				crc = (crc ^ ETHER_CRC_POLY_LE);
   1017 		}
   1018 	}
   1019 
   1020 	return (crc);
   1021 }
   1022 #else
   1023 uint32_t
   1024 ether_crc32_le(const uint8_t *buf, size_t len)
   1025 {
   1026 	static const uint32_t crctab[] = {
   1027 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
   1028 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
   1029 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
   1030 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
   1031 	};
   1032 	uint32_t crc;
   1033 	size_t i;
   1034 
   1035 	crc = 0xffffffffU;	/* initial value */
   1036 
   1037 	for (i = 0; i < len; i++) {
   1038 		crc ^= buf[i];
   1039 		crc = (crc >> 4) ^ crctab[crc & 0xf];
   1040 		crc = (crc >> 4) ^ crctab[crc & 0xf];
   1041 	}
   1042 
   1043 	return (crc);
   1044 }
   1045 #endif
   1046 
   1047 uint32_t
   1048 ether_crc32_be(const uint8_t *buf, size_t len)
   1049 {
   1050 	uint32_t c, crc, carry;
   1051 	size_t i, j;
   1052 
   1053 	crc = 0xffffffffU;	/* initial value */
   1054 
   1055 	for (i = 0; i < len; i++) {
   1056 		c = buf[i];
   1057 		for (j = 0; j < 8; j++) {
   1058 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
   1059 			crc <<= 1;
   1060 			c >>= 1;
   1061 			if (carry)
   1062 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
   1063 		}
   1064 	}
   1065 
   1066 	return (crc);
   1067 }
   1068 
   1069 #ifdef INET
   1070 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] =
   1071     { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 };
   1072 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] =
   1073     { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff };
   1074 #endif
   1075 #ifdef INET6
   1076 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] =
   1077     { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 };
   1078 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] =
   1079     { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff };
   1080 #endif
   1081 
   1082 /*
   1083  * ether_aton implementation, not using a static buffer.
   1084  */
   1085 int
   1086 ether_aton_r(u_char *dest, size_t len, const char *str)
   1087 {
   1088         const u_char *cp = (const void *)str;
   1089 	u_char *ep;
   1090 
   1091 #define atox(c)	(((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10))
   1092 
   1093 	if (len < ETHER_ADDR_LEN)
   1094 		return ENOSPC;
   1095 
   1096 	ep = dest + ETHER_ADDR_LEN;
   1097 
   1098 	while (*cp) {
   1099                 if (!isxdigit(*cp))
   1100                         return EINVAL;
   1101 		*dest = atox(*cp);
   1102 		cp++;
   1103                 if (isxdigit(*cp)) {
   1104                         *dest = (*dest << 4) | atox(*cp);
   1105 			dest++;
   1106 			cp++;
   1107                 } else
   1108 			dest++;
   1109 		if (dest == ep)
   1110 			return *cp == '\0' ? 0 : ENAMETOOLONG;
   1111 		switch (*cp) {
   1112 		case ':':
   1113 		case '-':
   1114 		case '.':
   1115 			cp++;
   1116 			break;
   1117 		}
   1118         }
   1119 	return ENOBUFS;
   1120 }
   1121 
   1122 /*
   1123  * Convert a sockaddr into an Ethernet address or range of Ethernet
   1124  * addresses.
   1125  */
   1126 int
   1127 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN],
   1128     uint8_t addrhi[ETHER_ADDR_LEN])
   1129 {
   1130 #ifdef INET
   1131 	const struct sockaddr_in *sin;
   1132 #endif /* INET */
   1133 #ifdef INET6
   1134 	const struct sockaddr_in6 *sin6;
   1135 #endif /* INET6 */
   1136 
   1137 	switch (sa->sa_family) {
   1138 
   1139 	case AF_UNSPEC:
   1140 		memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN);
   1141 		memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
   1142 		break;
   1143 
   1144 #ifdef INET
   1145 	case AF_INET:
   1146 		sin = satocsin(sa);
   1147 		if (sin->sin_addr.s_addr == INADDR_ANY) {
   1148 			/*
   1149 			 * An IP address of INADDR_ANY means listen to
   1150 			 * or stop listening to all of the Ethernet
   1151 			 * multicast addresses used for IP.
   1152 			 * (This is for the sake of IP multicast routers.)
   1153 			 */
   1154 			memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN);
   1155 			memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN);
   1156 		}
   1157 		else {
   1158 			ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo);
   1159 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
   1160 		}
   1161 		break;
   1162 #endif
   1163 #ifdef INET6
   1164 	case AF_INET6:
   1165 		sin6 = satocsin6(sa);
   1166 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
   1167 			/*
   1168 			 * An IP6 address of 0 means listen to or stop
   1169 			 * listening to all of the Ethernet multicast
   1170 			 * address used for IP6.
   1171 			 * (This is used for multicast routers.)
   1172 			 */
   1173 			memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN);
   1174 			memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN);
   1175 		} else {
   1176 			ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo);
   1177 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
   1178 		}
   1179 		break;
   1180 #endif
   1181 
   1182 	default:
   1183 		return EAFNOSUPPORT;
   1184 	}
   1185 	return 0;
   1186 }
   1187 
   1188 /*
   1189  * Add an Ethernet multicast address or range of addresses to the list for a
   1190  * given interface.
   1191  */
   1192 int
   1193 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec)
   1194 {
   1195 	struct ether_multi *enm;
   1196 	u_char addrlo[ETHER_ADDR_LEN];
   1197 	u_char addrhi[ETHER_ADDR_LEN];
   1198 	int s = splnet(), error;
   1199 
   1200 	error = ether_multiaddr(sa, addrlo, addrhi);
   1201 	if (error != 0) {
   1202 		splx(s);
   1203 		return error;
   1204 	}
   1205 
   1206 	/*
   1207 	 * Verify that we have valid Ethernet multicast addresses.
   1208 	 */
   1209 	if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) {
   1210 		splx(s);
   1211 		return EINVAL;
   1212 	}
   1213 	/*
   1214 	 * See if the address range is already in the list.
   1215 	 */
   1216 	ETHER_LOOKUP_MULTI(addrlo, addrhi, ec, enm);
   1217 	if (enm != NULL) {
   1218 		/*
   1219 		 * Found it; just increment the reference count.
   1220 		 */
   1221 		++enm->enm_refcount;
   1222 		splx(s);
   1223 		return 0;
   1224 	}
   1225 	/*
   1226 	 * New address or range; malloc a new multicast record
   1227 	 * and link it into the interface's multicast list.
   1228 	 */
   1229 	enm = (struct ether_multi *)malloc(sizeof(*enm), M_IFMADDR, M_NOWAIT);
   1230 	if (enm == NULL) {
   1231 		splx(s);
   1232 		return ENOBUFS;
   1233 	}
   1234 	memcpy(enm->enm_addrlo, addrlo, 6);
   1235 	memcpy(enm->enm_addrhi, addrhi, 6);
   1236 	enm->enm_refcount = 1;
   1237 	LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list);
   1238 	ec->ec_multicnt++;
   1239 	splx(s);
   1240 	/*
   1241 	 * Return ENETRESET to inform the driver that the list has changed
   1242 	 * and its reception filter should be adjusted accordingly.
   1243 	 */
   1244 	return ENETRESET;
   1245 }
   1246 
   1247 /*
   1248  * Delete a multicast address record.
   1249  */
   1250 int
   1251 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec)
   1252 {
   1253 	struct ether_multi *enm;
   1254 	u_char addrlo[ETHER_ADDR_LEN];
   1255 	u_char addrhi[ETHER_ADDR_LEN];
   1256 	int s = splnet(), error;
   1257 
   1258 	error = ether_multiaddr(sa, addrlo, addrhi);
   1259 	if (error != 0) {
   1260 		splx(s);
   1261 		return (error);
   1262 	}
   1263 
   1264 	/*
   1265 	 * Look ur the address in our list.
   1266 	 */
   1267 	ETHER_LOOKUP_MULTI(addrlo, addrhi, ec, enm);
   1268 	if (enm == NULL) {
   1269 		splx(s);
   1270 		return (ENXIO);
   1271 	}
   1272 	if (--enm->enm_refcount != 0) {
   1273 		/*
   1274 		 * Still some claims to this record.
   1275 		 */
   1276 		splx(s);
   1277 		return (0);
   1278 	}
   1279 	/*
   1280 	 * No remaining claims to this record; unlink and free it.
   1281 	 */
   1282 	LIST_REMOVE(enm, enm_list);
   1283 	free(enm, M_IFMADDR);
   1284 	ec->ec_multicnt--;
   1285 	splx(s);
   1286 	/*
   1287 	 * Return ENETRESET to inform the driver that the list has changed
   1288 	 * and its reception filter should be adjusted accordingly.
   1289 	 */
   1290 	return (ENETRESET);
   1291 }
   1292 
   1293 void
   1294 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb)
   1295 {
   1296 	ec->ec_ifflags_cb = cb;
   1297 }
   1298 
   1299 /*
   1300  * Common ioctls for Ethernet interfaces.  Note, we must be
   1301  * called at splnet().
   1302  */
   1303 int
   1304 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data)
   1305 {
   1306 	struct ethercom *ec = (void *) ifp;
   1307 	struct eccapreq *eccr;
   1308 	struct ifreq *ifr = (struct ifreq *)data;
   1309 	struct if_laddrreq *iflr = data;
   1310 	const struct sockaddr_dl *sdl;
   1311 	static const uint8_t zero[ETHER_ADDR_LEN];
   1312 	int error;
   1313 
   1314 	switch (cmd) {
   1315 	case SIOCINITIFADDR:
   1316 	    {
   1317 		struct ifaddr *ifa = (struct ifaddr *)data;
   1318 		if (ifa->ifa_addr->sa_family != AF_LINK
   1319 		    && (ifp->if_flags & (IFF_UP|IFF_RUNNING)) !=
   1320 		       (IFF_UP|IFF_RUNNING)) {
   1321 			ifp->if_flags |= IFF_UP;
   1322 			if ((error = (*ifp->if_init)(ifp)) != 0)
   1323 				return error;
   1324 		}
   1325 #ifdef INET
   1326 		if (ifa->ifa_addr->sa_family == AF_INET)
   1327 			arp_ifinit(ifp, ifa);
   1328 #endif /* INET */
   1329 		return 0;
   1330 	    }
   1331 
   1332 	case SIOCSIFMTU:
   1333 	    {
   1334 		int maxmtu;
   1335 
   1336 		if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU)
   1337 			maxmtu = ETHERMTU_JUMBO;
   1338 		else
   1339 			maxmtu = ETHERMTU;
   1340 
   1341 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu)
   1342 			return EINVAL;
   1343 		else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET)
   1344 			return error;
   1345 		else if (ifp->if_flags & IFF_UP) {
   1346 			/* Make sure the device notices the MTU change. */
   1347 			return (*ifp->if_init)(ifp);
   1348 		} else
   1349 			return 0;
   1350 	    }
   1351 
   1352 	case SIOCSIFFLAGS:
   1353 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
   1354 			return error;
   1355 		switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
   1356 		case IFF_RUNNING:
   1357 			/*
   1358 			 * If interface is marked down and it is running,
   1359 			 * then stop and disable it.
   1360 			 */
   1361 			(*ifp->if_stop)(ifp, 1);
   1362 			break;
   1363 		case IFF_UP:
   1364 			/*
   1365 			 * If interface is marked up and it is stopped, then
   1366 			 * start it.
   1367 			 */
   1368 			return (*ifp->if_init)(ifp);
   1369 		case IFF_UP|IFF_RUNNING:
   1370 			error = 0;
   1371 			if (ec->ec_ifflags_cb == NULL ||
   1372 			    (error = (*ec->ec_ifflags_cb)(ec)) == ENETRESET) {
   1373 				/*
   1374 				 * Reset the interface to pick up
   1375 				 * changes in any other flags that
   1376 				 * affect the hardware state.
   1377 				 */
   1378 				return (*ifp->if_init)(ifp);
   1379 			} else
   1380 				return error;
   1381 		case 0:
   1382 			break;
   1383 		}
   1384 		return 0;
   1385 	case SIOCGETHERCAP:
   1386 		eccr = (struct eccapreq *)data;
   1387 		eccr->eccr_capabilities = ec->ec_capabilities;
   1388 		eccr->eccr_capenable = ec->ec_capenable;
   1389 		return 0;
   1390 	case SIOCADDMULTI:
   1391 		return ether_addmulti(ifreq_getaddr(cmd, ifr), ec);
   1392 	case SIOCDELMULTI:
   1393 		return ether_delmulti(ifreq_getaddr(cmd, ifr), ec);
   1394 	case SIOCSIFMEDIA:
   1395 	case SIOCGIFMEDIA:
   1396 		if (ec->ec_mii == NULL)
   1397 			return ENOTTY;
   1398 		return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media, cmd);
   1399 	case SIOCALIFADDR:
   1400 		sdl = satocsdl(sstocsa(&iflr->addr));
   1401 		if (sdl->sdl_family != AF_LINK)
   1402 			;
   1403 		else if (ETHER_IS_MULTICAST(CLLADDR(sdl)))
   1404 			return EINVAL;
   1405 		else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0)
   1406 			return EINVAL;
   1407 		/*FALLTHROUGH*/
   1408 	default:
   1409 		return ifioctl_common(ifp, cmd, data);
   1410 	}
   1411 	return 0;
   1412 }
   1413 
   1414 /*
   1415  * Enable/disable passing VLAN packets if the parent interface supports it.
   1416  * Return:
   1417  * 	 0: Ok
   1418  *	-1: Parent interface does not support vlans
   1419  *	>0: Error
   1420  */
   1421 int
   1422 ether_enable_vlan_mtu(struct ifnet *ifp)
   1423 {
   1424 	int error;
   1425 	struct ethercom *ec = (void *)ifp;
   1426 
   1427 	/* Already have VLAN's do nothing. */
   1428 	if (ec->ec_nvlans != 0)
   1429 		return 0;
   1430 
   1431 	/* Parent does not support VLAN's */
   1432 	if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0)
   1433 		return -1;
   1434 
   1435 	/*
   1436 	 * Parent supports the VLAN_MTU capability,
   1437 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames;
   1438 	 * enable it.
   1439 	 */
   1440 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
   1441 
   1442 	/* Interface is down, defer for later */
   1443 	if ((ifp->if_flags & IFF_UP) == 0)
   1444 		return 0;
   1445 
   1446 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
   1447 		return 0;
   1448 
   1449 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
   1450 	return error;
   1451 }
   1452 
   1453 int
   1454 ether_disable_vlan_mtu(struct ifnet *ifp)
   1455 {
   1456 	int error;
   1457 	struct ethercom *ec = (void *)ifp;
   1458 
   1459 	/* We still have VLAN's, defer for later */
   1460 	if (ec->ec_nvlans != 0)
   1461 		return 0;
   1462 
   1463 	/* Parent does not support VLAB's, nothing to do. */
   1464 	if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0)
   1465 		return -1;
   1466 
   1467 	/*
   1468 	 * Disable Tx/Rx of VLAN-sized frames.
   1469 	 */
   1470 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
   1471 
   1472 	/* Interface is down, defer for later */
   1473 	if ((ifp->if_flags & IFF_UP) == 0)
   1474 		return 0;
   1475 
   1476 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
   1477 		return 0;
   1478 
   1479 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
   1480 	return error;
   1481 }
   1482 
   1483 static int
   1484 ether_multicast_sysctl(SYSCTLFN_ARGS)
   1485 {
   1486 	struct ether_multi *enm;
   1487 	struct ether_multi_sysctl addr;
   1488 	struct ifnet *ifp;
   1489 	struct ethercom *ec;
   1490 	int error = 0;
   1491 	size_t written;
   1492 	struct psref psref;
   1493 	int bound;
   1494 
   1495 	if (namelen != 1)
   1496 		return EINVAL;
   1497 
   1498 	bound = curlwp_bind();
   1499 	ifp = if_get_byindex(name[0], &psref);
   1500 	if (ifp == NULL) {
   1501 		error = ENODEV;
   1502 		goto out;
   1503 	}
   1504 	if (ifp->if_type != IFT_ETHER) {
   1505 		if_put(ifp, &psref);
   1506 		*oldlenp = 0;
   1507 		goto out;
   1508 	}
   1509 	ec = (struct ethercom *)ifp;
   1510 
   1511 	if (oldp == NULL) {
   1512 		if_put(ifp, &psref);
   1513 		*oldlenp = ec->ec_multicnt * sizeof(addr);
   1514 		goto out;
   1515 	}
   1516 
   1517 	memset(&addr, 0, sizeof(addr));
   1518 	error = 0;
   1519 	written = 0;
   1520 
   1521 	LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) {
   1522 		if (written + sizeof(addr) > *oldlenp)
   1523 			break;
   1524 		addr.enm_refcount = enm->enm_refcount;
   1525 		memcpy(addr.enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN);
   1526 		memcpy(addr.enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN);
   1527 		error = sysctl_copyout(l, &addr, oldp, sizeof(addr));
   1528 		if (error)
   1529 			break;
   1530 		written += sizeof(addr);
   1531 		oldp = (char *)oldp + sizeof(addr);
   1532 	}
   1533 	if_put(ifp, &psref);
   1534 
   1535 	*oldlenp = written;
   1536 out:
   1537 	curlwp_bindx(bound);
   1538 	return error;
   1539 }
   1540 
   1541 SYSCTL_SETUP(sysctl_net_ether_setup, "sysctl net.ether subtree setup")
   1542 {
   1543 	const struct sysctlnode *rnode = NULL;
   1544 
   1545 	sysctl_createv(clog, 0, NULL, &rnode,
   1546 		       CTLFLAG_PERMANENT,
   1547 		       CTLTYPE_NODE, "ether",
   1548 		       SYSCTL_DESCR("Ethernet-specific information"),
   1549 		       NULL, 0, NULL, 0,
   1550 		       CTL_NET, CTL_CREATE, CTL_EOL);
   1551 
   1552 	sysctl_createv(clog, 0, &rnode, NULL,
   1553 		       CTLFLAG_PERMANENT,
   1554 		       CTLTYPE_NODE, "multicast",
   1555 		       SYSCTL_DESCR("multicast addresses"),
   1556 		       ether_multicast_sysctl, 0, NULL, 0,
   1557 		       CTL_CREATE, CTL_EOL);
   1558 }
   1559 
   1560 void
   1561 etherinit(void)
   1562 {
   1563 	mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET);
   1564 }
   1565