Home | History | Annotate | Line # | Download | only in net
      1 /*	$NetBSD: if_ethersubr.c,v 1.333 2025/10/12 23:41:09 thorpej 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 1.333 2025/10/12 23:41:09 thorpej Exp $");
     65 
     66 #ifdef _KERNEL_OPT
     67 #include "opt_inet.h"
     68 #include "opt_atalk.h"
     69 #include "opt_mbuftrace.h"
     70 #include "opt_mpls.h"
     71 #include "opt_gateway.h"
     72 #include "opt_pppoe.h"
     73 #include "opt_net_mpsafe.h"
     74 #endif
     75 
     76 #include "vlan.h"
     77 #include "pppoe.h"
     78 #include "bridge.h"
     79 #include "arp.h"
     80 #include "agr.h"
     81 
     82 #include <sys/sysctl.h>
     83 #include <sys/mbuf.h>
     84 #include <sys/mutex.h>
     85 #include <sys/ioctl.h>
     86 #include <sys/errno.h>
     87 #include <sys/device.h>
     88 #include <sys/entropy.h>
     89 #include <sys/rndsource.h>
     90 #include <sys/cpu.h>
     91 #include <sys/kmem.h>
     92 #include <sys/hook.h>
     93 
     94 #include <net/if.h>
     95 #include <net/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 #include <net/ether_calls.h>
    117 
    118 #if NPPPOE > 0
    119 #include <net/if_pppoe.h>
    120 #endif
    121 
    122 #if NAGR > 0
    123 #include <net/ether_slowprotocols.h>
    124 #include <net/agr/ieee8023ad.h>
    125 #include <net/agr/if_agrvar.h>
    126 #endif
    127 
    128 #if NBRIDGE > 0
    129 #include <net/if_bridgevar.h>
    130 #endif
    131 
    132 #include <netinet/in.h>
    133 #ifdef INET
    134 #include <netinet/in_var.h>
    135 #endif
    136 #include <netinet/if_inarp.h>
    137 
    138 #ifdef INET6
    139 #ifndef INET
    140 #include <netinet/in.h>
    141 #endif
    142 #include <netinet6/in6_var.h>
    143 #include <netinet6/nd6.h>
    144 #endif
    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 CTASSERT(sizeof(struct ether_addr) == 6);
    169 CTASSERT(sizeof(struct ether_header) == 14);
    170 
    171 #ifdef DIAGNOSTIC
    172 static struct timeval bigpktppslim_last;
    173 static int bigpktppslim = 2;	/* XXX */
    174 static int bigpktpps_count;
    175 static kmutex_t bigpktpps_lock __cacheline_aligned;
    176 #endif
    177 
    178 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] =
    179     { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
    180 const uint8_t ethermulticastaddr_slowprotocols[ETHER_ADDR_LEN] =
    181     { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x02 };
    182 #define senderr(e) { error = (e); goto bad;}
    183 
    184 static pktq_rps_hash_func_t ether_pktq_rps_hash_p;
    185 
    186 /*
    187  * Attempt to get the Ethernet address from device properties.
    188  *
    189  * We generally follow the Device Tree specification with regard
    190  * to the property names, but that's OK because the property names
    191  * are pretty generic.
    192  */
    193 bool
    194 ether_getaddr(device_t dev, uint8_t enaddr[ETHER_ADDR_LEN])
    195 {
    196 	/*
    197 	 * First check the platform device tree; it may have specific
    198 	 * rules about how Ethernet addresses are assigned, separate
    199 	 * from properties associated with a given device.
    200 	 */
    201 	struct ether_get_mac_address_args args = {
    202 		.enaddr = enaddr,
    203 	};
    204 	if (device_call(dev, ETHER_GET_MAC_ADDRESS(&args)) == 0) {
    205 		/* Got it from the platform device tree. */
    206 		return true;
    207 	}
    208 
    209 	/*
    210 	 * Check first for the "mac-address" property.  The bindings
    211 	 * say that this would be used only if it is different then the
    212 	 * "local-mac-address" property.  But if it's the same, then
    213 	 * using it is exactly the same as using "local-mac-address". So,
    214 	 * we first look for "mac-address", and if that's not there, then
    215 	 * we look for "local-mac-address".
    216 	 */
    217 	if (device_getprop_data(dev, "mac-address", enaddr,
    218 				ETHER_ADDR_LEN) == ETHER_ADDR_LEN ||
    219 	    device_getprop_data(dev, "local-mac-address", enaddr,
    220 				ETHER_ADDR_LEN) == ETHER_ADDR_LEN) {
    221 		return true;
    222 	}
    223 	return false;
    224 }
    225 
    226 /*
    227  * Ethernet output routine.
    228  * Encapsulate a packet of type family for the local net.
    229  * Assumes that ifp is actually pointer to ethercom structure.
    230  */
    231 static int
    232 ether_output(struct ifnet * const ifp0, struct mbuf * const m0,
    233     const struct sockaddr * const dst, const struct rtentry *rt)
    234 {
    235 	uint8_t esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN];
    236 	uint16_t etype = 0;
    237 	int error = 0, hdrcmplt = 0;
    238 	struct mbuf *m = m0;
    239 	struct mbuf *mcopy = NULL;
    240 	struct ether_header *eh;
    241 	struct ifnet *ifp = ifp0;
    242 #ifdef INET
    243 	struct arphdr *ah;
    244 #endif
    245 #ifdef NETATALK
    246 	struct at_ifaddr *aa;
    247 #endif
    248 
    249 #ifdef MBUFTRACE
    250 	m_claimm(m, ifp->if_mowner);
    251 #endif
    252 
    253 #if NCARP > 0
    254 	if (ifp->if_type == IFT_CARP) {
    255 		struct ifaddr *ifa;
    256 		int s = pserialize_read_enter();
    257 
    258 		/* loop back if this is going to the carp interface */
    259 		if (dst != NULL && ifp0->if_link_state == LINK_STATE_UP &&
    260 		    (ifa = ifa_ifwithaddr(dst)) != NULL) {
    261 			if (ifa->ifa_ifp == ifp0) {
    262 				pserialize_read_exit(s);
    263 				return looutput(ifp0, m, dst, rt);
    264 			}
    265 		}
    266 		pserialize_read_exit(s);
    267 
    268 		ifp = ifp->if_carpdev;
    269 		/* ac = (struct arpcom *)ifp; */
    270 
    271 		if ((ifp0->if_flags & (IFF_UP | IFF_RUNNING)) !=
    272 		    (IFF_UP | IFF_RUNNING))
    273 			senderr(ENETDOWN);
    274 	}
    275 #endif
    276 
    277 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
    278 		senderr(ENETDOWN);
    279 
    280 	switch (dst->sa_family) {
    281 
    282 #ifdef INET
    283 	case AF_INET:
    284 		if (m->m_flags & M_BCAST) {
    285 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
    286 		} else if (m->m_flags & M_MCAST) {
    287 			ETHER_MAP_IP_MULTICAST(&satocsin(dst)->sin_addr, edst);
    288 		} else {
    289 			error = arpresolve(ifp0, rt, m, dst, edst, sizeof(edst));
    290 			if (error)
    291 				return (error == EWOULDBLOCK) ? 0 : error;
    292 		}
    293 		/* If broadcasting on a simplex interface, loopback a copy */
    294 		if ((m->m_flags & M_BCAST) && (ifp->if_flags & IFF_SIMPLEX))
    295 			mcopy = m_copypacket(m, M_DONTWAIT);
    296 		etype = htons(ETHERTYPE_IP);
    297 		break;
    298 
    299 	case AF_ARP:
    300 		ah = mtod(m, struct arphdr *);
    301 		if (m->m_flags & M_BCAST) {
    302 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
    303 		} else {
    304 			void *tha = ar_tha(ah);
    305 
    306 			if (tha == NULL) {
    307 				/* fake with ARPHRD_IEEE1394 */
    308 				m_freem(m);
    309 				return 0;
    310 			}
    311 			memcpy(edst, tha, sizeof(edst));
    312 		}
    313 
    314 		ah->ar_hrd = htons(ARPHRD_ETHER);
    315 
    316 		switch (ntohs(ah->ar_op)) {
    317 		case ARPOP_REVREQUEST:
    318 		case ARPOP_REVREPLY:
    319 			etype = htons(ETHERTYPE_REVARP);
    320 			break;
    321 
    322 		case ARPOP_REQUEST:
    323 		case ARPOP_REPLY:
    324 		default:
    325 			etype = htons(ETHERTYPE_ARP);
    326 		}
    327 		break;
    328 #endif
    329 
    330 #ifdef INET6
    331 	case AF_INET6:
    332 		if (m->m_flags & M_BCAST) {
    333 			memcpy(edst, etherbroadcastaddr, sizeof(edst));
    334 		} else if (m->m_flags & M_MCAST) {
    335 			ETHER_MAP_IPV6_MULTICAST(&satocsin6(dst)->sin6_addr,
    336 			    edst);
    337 		} else {
    338 			error = nd6_resolve(ifp0, rt, m, dst, edst,
    339 			    sizeof(edst));
    340 			if (error)
    341 				return (error == EWOULDBLOCK) ? 0 : error;
    342 		}
    343 		etype = htons(ETHERTYPE_IPV6);
    344 		break;
    345 #endif
    346 
    347 #ifdef NETATALK
    348 	case AF_APPLETALK: {
    349 		struct ifaddr *ifa;
    350 		int s;
    351 
    352 		KERNEL_LOCK(1, NULL);
    353 
    354 		if (!aarpresolve(ifp, m, (const struct sockaddr_at *)dst, edst)) {
    355 			KERNEL_UNLOCK_ONE(NULL);
    356 			return 0;
    357 		}
    358 
    359 		/*
    360 		 * ifaddr is the first thing in at_ifaddr
    361 		 */
    362 		s = pserialize_read_enter();
    363 		ifa = at_ifawithnet((const struct sockaddr_at *)dst, ifp);
    364 		if (ifa == NULL) {
    365 			pserialize_read_exit(s);
    366 			KERNEL_UNLOCK_ONE(NULL);
    367 			senderr(EADDRNOTAVAIL);
    368 		}
    369 		aa = (struct at_ifaddr *)ifa;
    370 
    371 		/*
    372 		 * In the phase 2 case, we need to prepend an mbuf for the
    373 		 * llc header.
    374 		 */
    375 		if (aa->aa_flags & AFA_PHASE2) {
    376 			struct llc llc;
    377 
    378 			M_PREPEND(m, sizeof(struct llc), M_DONTWAIT);
    379 			if (m == NULL) {
    380 				pserialize_read_exit(s);
    381 				KERNEL_UNLOCK_ONE(NULL);
    382 				senderr(ENOBUFS);
    383 			}
    384 
    385 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
    386 			llc.llc_control = LLC_UI;
    387 			memcpy(llc.llc_snap_org_code, at_org_code,
    388 			    sizeof(llc.llc_snap_org_code));
    389 			llc.llc_snap_ether_type = htons(ETHERTYPE_ATALK);
    390 			memcpy(mtod(m, void *), &llc, sizeof(struct llc));
    391 		} else {
    392 			etype = htons(ETHERTYPE_ATALK);
    393 		}
    394 		pserialize_read_exit(s);
    395 		KERNEL_UNLOCK_ONE(NULL);
    396 		break;
    397 	}
    398 #endif /* NETATALK */
    399 
    400 	case pseudo_AF_HDRCMPLT:
    401 		hdrcmplt = 1;
    402 		memcpy(esrc,
    403 		    ((const struct ether_header *)dst->sa_data)->ether_shost,
    404 		    sizeof(esrc));
    405 		/* FALLTHROUGH */
    406 
    407 	case AF_UNSPEC:
    408 		memcpy(edst,
    409 		    ((const struct ether_header *)dst->sa_data)->ether_dhost,
    410 		    sizeof(edst));
    411 		/* AF_UNSPEC doesn't swap the byte order of the ether_type. */
    412 		etype = ((const struct ether_header *)dst->sa_data)->ether_type;
    413 		break;
    414 
    415 	default:
    416 		rt_unhandled(__func__, ifp, dst);
    417 		senderr(EAFNOSUPPORT);
    418 	}
    419 
    420 #ifdef MPLS
    421 	{
    422 		struct m_tag *mtag;
    423 		mtag = m_tag_find(m, PACKET_TAG_MPLS);
    424 		if (mtag != NULL) {
    425 			/* Having the tag itself indicates it's MPLS */
    426 			etype = htons(ETHERTYPE_MPLS);
    427 			m_tag_delete(m, mtag);
    428 		}
    429 	}
    430 #endif
    431 
    432 	if (mcopy)
    433 		(void)looutput(ifp, mcopy, dst, rt);
    434 
    435 	KASSERT((m->m_flags & M_PKTHDR) != 0);
    436 
    437 	/*
    438 	 * If no ether type is set, this must be a 802.2 formatted packet.
    439 	 */
    440 	if (etype == 0)
    441 		etype = htons(m->m_pkthdr.len);
    442 
    443 	/*
    444 	 * Add local net header. If no space in first mbuf, allocate another.
    445 	 */
    446 	M_PREPEND(m, sizeof(struct ether_header), M_DONTWAIT);
    447 	if (m == NULL)
    448 		senderr(ENOBUFS);
    449 
    450 	eh = mtod(m, struct ether_header *);
    451 	/* Note: etype is already in network byte order. */
    452 	memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
    453 	memcpy(eh->ether_dhost, edst, sizeof(edst));
    454 	if (hdrcmplt) {
    455 		memcpy(eh->ether_shost, esrc, sizeof(eh->ether_shost));
    456 	} else {
    457 	 	memcpy(eh->ether_shost, CLLADDR(ifp->if_sadl),
    458 		    sizeof(eh->ether_shost));
    459 	}
    460 
    461 #if NCARP > 0
    462 	if (ifp0 != ifp && ifp0->if_type == IFT_CARP) {
    463 		/* update with virtual MAC */
    464 		memcpy(eh->ether_shost, CLLADDR(ifp0->if_sadl),
    465 		    sizeof(eh->ether_shost));
    466 	}
    467 #endif
    468 
    469 	if ((error = pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_OUT)) != 0)
    470 		return error;
    471 	if (m == NULL)
    472 		return 0;
    473 
    474 #if NBRIDGE > 0
    475 	/*
    476 	 * Bridges require special output handling.
    477 	 */
    478 	if (ifp->if_bridge)
    479 		return bridge_output(ifp, m, NULL, NULL);
    480 #endif
    481 
    482 #if NCARP > 0
    483 	if (ifp != ifp0)
    484 		if_statadd(ifp0, if_obytes, m->m_pkthdr.len + ETHER_HDR_LEN);
    485 #endif
    486 
    487 #ifdef ALTQ
    488 	KERNEL_LOCK(1, NULL);
    489 	/*
    490 	 * If ALTQ is enabled on the parent interface, do
    491 	 * classification; the queueing discipline might not
    492 	 * require classification, but might require the
    493 	 * address family/header pointer in the pktattr.
    494 	 */
    495 	if (ALTQ_IS_ENABLED(&ifp->if_snd))
    496 		altq_etherclassify(&ifp->if_snd, m);
    497 	KERNEL_UNLOCK_ONE(NULL);
    498 #endif
    499 	return ifq_enqueue(ifp, m);
    500 
    501 bad:
    502 	if_statinc(ifp, if_oerrors);
    503 	m_freem(m);
    504 	return error;
    505 }
    506 
    507 #ifdef ALTQ
    508 /*
    509  * This routine is a slight hack to allow a packet to be classified
    510  * if the Ethernet headers are present.  It will go away when ALTQ's
    511  * classification engine understands link headers.
    512  *
    513  * XXX: We may need to do m_pullups here. First to ensure struct ether_header
    514  * is indeed contiguous, then to read the LLC and so on.
    515  */
    516 void
    517 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m)
    518 {
    519 	struct ether_header *eh;
    520 	struct mbuf *mtop = m;
    521 	uint16_t ether_type;
    522 	int hlen, af, hdrsize;
    523 	void *hdr;
    524 
    525 	KASSERT((mtop->m_flags & M_PKTHDR) != 0);
    526 
    527 	hlen = ETHER_HDR_LEN;
    528 	eh = mtod(m, struct ether_header *);
    529 
    530 	ether_type = htons(eh->ether_type);
    531 
    532 	if (ether_type < ETHERMTU) {
    533 		/* LLC/SNAP */
    534 		struct llc *llc = (struct llc *)(eh + 1);
    535 		hlen += 8;
    536 
    537 		if (m->m_len < hlen ||
    538 		    llc->llc_dsap != LLC_SNAP_LSAP ||
    539 		    llc->llc_ssap != LLC_SNAP_LSAP ||
    540 		    llc->llc_control != LLC_UI) {
    541 			/* Not SNAP. */
    542 			goto bad;
    543 		}
    544 
    545 		ether_type = htons(llc->llc_un.type_snap.ether_type);
    546 	}
    547 
    548 	switch (ether_type) {
    549 	case ETHERTYPE_IP:
    550 		af = AF_INET;
    551 		hdrsize = 20;		/* sizeof(struct ip) */
    552 		break;
    553 
    554 	case ETHERTYPE_IPV6:
    555 		af = AF_INET6;
    556 		hdrsize = 40;		/* sizeof(struct ip6_hdr) */
    557 		break;
    558 
    559 	default:
    560 		af = AF_UNSPEC;
    561 		hdrsize = 0;
    562 		break;
    563 	}
    564 
    565 	while (m->m_len <= hlen) {
    566 		hlen -= m->m_len;
    567 		m = m->m_next;
    568 		if (m == NULL)
    569 			goto bad;
    570 	}
    571 
    572 	if (m->m_len < (hlen + hdrsize)) {
    573 		/*
    574 		 * protocol header not in a single mbuf.
    575 		 * We can't cope with this situation right
    576 		 * now (but it shouldn't ever happen, really, anyhow).
    577 		 */
    578 #ifdef DEBUG
    579 		printf("altq_etherclassify: headers span multiple mbufs: "
    580 		    "%d < %d\n", m->m_len, (hlen + hdrsize));
    581 #endif
    582 		goto bad;
    583 	}
    584 
    585 	m->m_data += hlen;
    586 	m->m_len -= hlen;
    587 
    588 	hdr = mtod(m, void *);
    589 
    590 	if (ALTQ_NEEDS_CLASSIFY(ifq)) {
    591 		mtop->m_pkthdr.pattr_class =
    592 		    (*ifq->altq_classify)(ifq->altq_clfier, m, af);
    593 	}
    594 	mtop->m_pkthdr.pattr_af = af;
    595 	mtop->m_pkthdr.pattr_hdr = hdr;
    596 
    597 	m->m_data -= hlen;
    598 	m->m_len += hlen;
    599 
    600 	return;
    601 
    602 bad:
    603 	mtop->m_pkthdr.pattr_class = NULL;
    604 	mtop->m_pkthdr.pattr_hdr = NULL;
    605 	mtop->m_pkthdr.pattr_af = AF_UNSPEC;
    606 }
    607 #endif /* ALTQ */
    608 
    609 #if defined (LLC) || defined (NETATALK)
    610 static void
    611 ether_input_llc(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh)
    612 {
    613 	pktqueue_t *pktq = NULL;
    614 	struct llc *l;
    615 
    616 	if (m->m_len < sizeof(*eh) + sizeof(struct llc))
    617 		goto error;
    618 
    619 	l = (struct llc *)(eh+1);
    620 	switch (l->llc_dsap) {
    621 #ifdef NETATALK
    622 	case LLC_SNAP_LSAP:
    623 		switch (l->llc_control) {
    624 		case LLC_UI:
    625 			if (l->llc_ssap != LLC_SNAP_LSAP)
    626 				goto error;
    627 
    628 			if (memcmp(&(l->llc_snap_org_code)[0],
    629 			    at_org_code, sizeof(at_org_code)) == 0 &&
    630 			    ntohs(l->llc_snap_ether_type) ==
    631 			    ETHERTYPE_ATALK) {
    632 				pktq = at_pktq2;
    633 				m_adj(m, sizeof(struct ether_header)
    634 				    + sizeof(struct llc));
    635 				break;
    636 			}
    637 
    638 			if (memcmp(&(l->llc_snap_org_code)[0],
    639 			    aarp_org_code,
    640 			    sizeof(aarp_org_code)) == 0 &&
    641 			    ntohs(l->llc_snap_ether_type) ==
    642 			    ETHERTYPE_AARP) {
    643 				m_adj(m, sizeof(struct ether_header)
    644 				    + sizeof(struct llc));
    645 				aarpinput(ifp, m); /* XXX queue? */
    646 				return;
    647 			}
    648 
    649 		default:
    650 			goto error;
    651 		}
    652 		break;
    653 #endif
    654 	default:
    655 		goto noproto;
    656 	}
    657 
    658 	KASSERT(pktq != NULL);
    659 	if (__predict_false(!pktq_enqueue(pktq, m, 0))) {
    660 		m_freem(m);
    661 	}
    662 	return;
    663 
    664 noproto:
    665 	m_freem(m);
    666 	if_statinc(ifp, if_noproto);
    667 	return;
    668 error:
    669 	m_freem(m);
    670 	if_statinc(ifp, if_ierrors);
    671 	return;
    672 }
    673 #endif /* defined (LLC) || defined (NETATALK) */
    674 
    675 /*
    676  * Process a received Ethernet packet;
    677  * the packet is in the mbuf chain m with
    678  * the ether header.
    679  */
    680 void
    681 ether_input(struct ifnet *ifp, struct mbuf *m)
    682 {
    683 #if NVLAN > 0 || defined(MBUFTRACE)
    684 	struct ethercom *ec = (struct ethercom *) ifp;
    685 #endif
    686 	pktqueue_t *pktq = NULL;
    687 	uint16_t etype;
    688 	struct ether_header *eh;
    689 	size_t ehlen;
    690 	static int earlypkts;
    691 
    692 	/* No RPS for not-IP. */
    693 	pktq_rps_hash_func_t rps_hash = NULL;
    694 
    695 	KASSERT(!cpu_intr_p());
    696 	KASSERT((m->m_flags & M_PKTHDR) != 0);
    697 
    698 	if ((ifp->if_flags & IFF_UP) == 0)
    699 		goto drop;
    700 
    701 #ifdef MBUFTRACE
    702 	m_claimm(m, &ec->ec_rx_mowner);
    703 #endif
    704 
    705 	if (__predict_false(m->m_len < sizeof(*eh))) {
    706 		if ((m = m_pullup(m, sizeof(*eh))) == NULL) {
    707 			if_statinc(ifp, if_ierrors);
    708 			return;
    709 		}
    710 	}
    711 
    712 	eh = mtod(m, struct ether_header *);
    713 	etype = ntohs(eh->ether_type);
    714 	ehlen = sizeof(*eh);
    715 
    716 	if (__predict_false(earlypkts < 100 ||
    717 		entropy_epoch() == (unsigned)-1)) {
    718 		rnd_add_data(NULL, eh, ehlen, 0);
    719 		earlypkts++;
    720 	}
    721 
    722 	/*
    723 	 * Determine if the packet is within its size limits. For MPLS the
    724 	 * header length is variable, so we skip the check.
    725 	 */
    726 	if (etype != ETHERTYPE_MPLS && m->m_pkthdr.len >
    727 	    ETHER_MAX_FRAME(ifp, etype, m->m_flags & M_HASFCS)) {
    728 #ifdef DIAGNOSTIC
    729 		mutex_enter(&bigpktpps_lock);
    730 		if (ppsratecheck(&bigpktppslim_last, &bigpktpps_count,
    731 		    bigpktppslim)) {
    732 			printf("%s: discarding oversize frame (len=%d)\n",
    733 			    ifp->if_xname, m->m_pkthdr.len);
    734 		}
    735 		mutex_exit(&bigpktpps_lock);
    736 #endif
    737 		goto error;
    738 	}
    739 
    740 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
    741 		/*
    742 		 * If this is not a simplex interface, drop the packet
    743 		 * if it came from us.
    744 		 */
    745 		if ((ifp->if_flags & IFF_SIMPLEX) == 0 &&
    746 		    memcmp(CLLADDR(ifp->if_sadl), eh->ether_shost,
    747 		    ETHER_ADDR_LEN) == 0) {
    748 			goto drop;
    749 		}
    750 
    751 		if (memcmp(etherbroadcastaddr,
    752 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
    753 			m->m_flags |= M_BCAST;
    754 		else
    755 			m->m_flags |= M_MCAST;
    756 		if_statinc(ifp, if_imcasts);
    757 	}
    758 
    759 	/* If the CRC is still on the packet, trim it off. */
    760 	if (m->m_flags & M_HASFCS) {
    761 		m_adj(m, -ETHER_CRC_LEN);
    762 		m->m_flags &= ~M_HASFCS;
    763 	}
    764 
    765 	if_statadd(ifp, if_ibytes, m->m_pkthdr.len);
    766 
    767 	if (!vlan_has_tag(m) && etype == ETHERTYPE_VLAN) {
    768 		m = ether_strip_vlantag(m);
    769 		if (m == NULL) {
    770 			if_statinc(ifp, if_ierrors);
    771 			return;
    772 		}
    773 
    774 		eh = mtod(m, struct ether_header *);
    775 		etype = ntohs(eh->ether_type);
    776 		ehlen = sizeof(*eh);
    777 	}
    778 
    779 	if ((m->m_flags & (M_BCAST | M_MCAST | M_PROMISC)) == 0 &&
    780 	    (ifp->if_flags & IFF_PROMISC) != 0 &&
    781 	    memcmp(CLLADDR(ifp->if_sadl), eh->ether_dhost,
    782 	     ETHER_ADDR_LEN) != 0) {
    783 		m->m_flags |= M_PROMISC;
    784 	}
    785 
    786 	if ((m->m_flags & M_PROMISC) == 0) {
    787 		if (pfil_run_hooks(ifp->if_pfil, &m, ifp, PFIL_IN) != 0)
    788 			return;
    789 		if (m == NULL)
    790 			return;
    791 
    792 		eh = mtod(m, struct ether_header *);
    793 		etype = ntohs(eh->ether_type);
    794 	}
    795 
    796 	/*
    797 	 * Processing a logical interfaces that are able
    798 	 * to configure vlan(4).
    799 	*/
    800 #if NAGR > 0
    801 	if (ifp->if_lagg != NULL &&
    802 	    __predict_true(etype != ETHERTYPE_SLOWPROTOCOLS)) {
    803 		m->m_flags &= ~M_PROMISC;
    804 		agr_input(ifp, m);
    805 		return;
    806 	}
    807 #endif
    808 
    809 	/*
    810 	 * VLAN processing.
    811 	 *
    812 	 * VLAN provides service delimiting so the frames are
    813 	 * processed before other handlings. If a VLAN interface
    814 	 * does not exist to take those frames, they're returned
    815 	 * to ether_input().
    816 	 */
    817 
    818 	if (vlan_has_tag(m)) {
    819 		if (EVL_VLANOFTAG(vlan_get_tag(m)) == 0) {
    820 			if (etype == ETHERTYPE_VLAN ||
    821 			     etype == ETHERTYPE_QINQ)
    822 				goto drop;
    823 
    824 			/* XXX we should actually use the prio value? */
    825 			m->m_flags &= ~M_VLANTAG;
    826 		} else {
    827 #if NVLAN > 0
    828 			if (ec->ec_nvlans > 0) {
    829 				m = vlan_input(ifp, m);
    830 
    831 				/* vlan_input() called ether_input() recursively */
    832 				if (m == NULL)
    833 					return;
    834 			}
    835 #endif
    836 			/* drop VLAN frames not for this port. */
    837 			goto noproto;
    838 		}
    839 	}
    840 
    841 #if NCARP > 0
    842 	if (__predict_false(ifp->if_carp && ifp->if_type != IFT_CARP)) {
    843 		/*
    844 		 * Clear M_PROMISC, in case the packet comes from a
    845 		 * vlan.
    846 		 */
    847 		m->m_flags &= ~M_PROMISC;
    848 		if (carp_input(m, (uint8_t *)&eh->ether_shost,
    849 		    (uint8_t *)&eh->ether_dhost, eh->ether_type) == 0)
    850 			return;
    851 	}
    852 #endif
    853 
    854 	/*
    855 	 * Handle protocols that expect to have the Ethernet header
    856 	 * (and possibly FCS) intact.
    857 	 */
    858 	switch (etype) {
    859 #if NPPPOE > 0
    860 	case ETHERTYPE_PPPOEDISC:
    861 		pppoedisc_input(ifp, m);
    862 		return;
    863 
    864 	case ETHERTYPE_PPPOE:
    865 		pppoe_input(ifp, m);
    866 		return;
    867 #endif
    868 
    869 	case ETHERTYPE_SLOWPROTOCOLS: {
    870 		uint8_t subtype;
    871 
    872 		if (m->m_pkthdr.len < sizeof(*eh) + sizeof(subtype))
    873 			goto error;
    874 
    875 		m_copydata(m, sizeof(*eh), sizeof(subtype), &subtype);
    876 		switch (subtype) {
    877 #if NAGR > 0
    878 		case SLOWPROTOCOLS_SUBTYPE_LACP:
    879 			if (ifp->if_lagg != NULL) {
    880 				ieee8023ad_lacp_input(ifp, m);
    881 				return;
    882 			}
    883 			break;
    884 
    885 		case SLOWPROTOCOLS_SUBTYPE_MARKER:
    886 			if (ifp->if_lagg != NULL) {
    887 				ieee8023ad_marker_input(ifp, m);
    888 				return;
    889 			}
    890 			break;
    891 #endif
    892 
    893 		default:
    894 			if (subtype == 0 || subtype > 10) {
    895 				/* illegal value */
    896 				goto noproto;
    897 			}
    898 			/* unknown subtype */
    899 			break;
    900 		}
    901 	}
    902 	/* FALLTHROUGH */
    903 	default:
    904 		if (m->m_flags & M_PROMISC)
    905 			goto drop;
    906 	}
    907 
    908 	/* If the CRC is still on the packet, trim it off. */
    909 	if (m->m_flags & M_HASFCS) {
    910 		m_adj(m, -ETHER_CRC_LEN);
    911 		m->m_flags &= ~M_HASFCS;
    912 	}
    913 
    914 	/* etype represents the size of the payload in this case */
    915 	if (etype <= ETHERMTU + sizeof(struct ether_header)) {
    916 		KASSERT(ehlen == sizeof(*eh));
    917 #if defined (LLC) || defined (NETATALK)
    918 		ether_input_llc(ifp, m, eh);
    919 		return;
    920 #else
    921 		/* ethertype of 0-1500 is regarded as noproto */
    922 		goto noproto;
    923 #endif
    924 	}
    925 
    926 	/* For ARP packets, store the source address so that
    927 	 * ARP DAD probes can be validated. */
    928 	if (etype == ETHERTYPE_ARP) {
    929 		struct m_tag *mtag;
    930 
    931 		mtag = m_tag_get(PACKET_TAG_ETHERNET_SRC, ETHER_ADDR_LEN,
    932 		    M_NOWAIT);
    933 		if (mtag != NULL) {
    934 			memcpy(mtag + 1, &eh->ether_shost, ETHER_ADDR_LEN);
    935 			m_tag_prepend(m, mtag);
    936 		}
    937 	}
    938 
    939 	/* Strip off the Ethernet header. */
    940 	m_adj(m, ehlen);
    941 
    942 	switch (etype) {
    943 #ifdef INET
    944 	case ETHERTYPE_IP:
    945 #ifdef GATEWAY
    946 		if (ipflow_fastforward(m))
    947 			return;
    948 #endif
    949 		pktq = ip_pktq;
    950 		rps_hash = atomic_load_relaxed(&ether_pktq_rps_hash_p);
    951 		break;
    952 
    953 	case ETHERTYPE_ARP:
    954 		pktq = arp_pktq;
    955 		break;
    956 
    957 	case ETHERTYPE_REVARP:
    958 		revarpinput(m);	/* XXX queue? */
    959 		return;
    960 #endif
    961 
    962 #ifdef INET6
    963 	case ETHERTYPE_IPV6:
    964 		if (__predict_false(!in6_present))
    965 			goto noproto;
    966 #ifdef GATEWAY
    967 		if (ip6flow_fastforward(&m))
    968 			return;
    969 #endif
    970 		pktq = ip6_pktq;
    971 		rps_hash = atomic_load_relaxed(&ether_pktq_rps_hash_p);
    972 		break;
    973 #endif
    974 
    975 #ifdef NETATALK
    976 	case ETHERTYPE_ATALK:
    977 		pktq = at_pktq1;
    978 		break;
    979 
    980 	case ETHERTYPE_AARP:
    981 		aarpinput(ifp, m); /* XXX queue? */
    982 		return;
    983 #endif
    984 
    985 #ifdef MPLS
    986 	case ETHERTYPE_MPLS:
    987 		pktq = mpls_pktq;
    988 		break;
    989 #endif
    990 
    991 	default:
    992 		goto noproto;
    993 	}
    994 
    995 	KASSERT(pktq != NULL);
    996 	const uint32_t h = rps_hash ? pktq_rps_hash(&rps_hash, m) : 0;
    997 	if (__predict_false(!pktq_enqueue(pktq, m, h))) {
    998 		m_freem(m);
    999 	}
   1000 	return;
   1001 
   1002 drop:
   1003 	m_freem(m);
   1004 	if_statinc(ifp, if_iqdrops);
   1005 	return;
   1006 noproto:
   1007 	m_freem(m);
   1008 	if_statinc(ifp, if_noproto);
   1009 	return;
   1010 error:
   1011 	m_freem(m);
   1012 	if_statinc(ifp, if_ierrors);
   1013 	return;
   1014 }
   1015 
   1016 static void
   1017 ether_bpf_mtap(struct bpf_if *bp, struct mbuf *m, u_int direction)
   1018 {
   1019 	struct ether_vlan_header evl;
   1020 	struct m_hdr mh, md;
   1021 
   1022 	KASSERT(bp != NULL);
   1023 
   1024 	if (!vlan_has_tag(m)) {
   1025 		bpf_mtap3(bp, m, direction);
   1026 		return;
   1027 	}
   1028 
   1029 	memcpy(&evl, mtod(m, char *), ETHER_HDR_LEN);
   1030 	evl.evl_proto = evl.evl_encap_proto;
   1031 	evl.evl_encap_proto = htons(ETHERTYPE_VLAN);
   1032 	evl.evl_tag = htons(vlan_get_tag(m));
   1033 
   1034 	md.mh_flags = 0;
   1035 	md.mh_data = m->m_data + ETHER_HDR_LEN;
   1036 	md.mh_len = m->m_len - ETHER_HDR_LEN;
   1037 	md.mh_next = m->m_next;
   1038 
   1039 	mh.mh_flags = 0;
   1040 	mh.mh_data = (char *)&evl;
   1041 	mh.mh_len = sizeof(evl);
   1042 	mh.mh_next = (struct mbuf *)&md;
   1043 
   1044 	bpf_mtap3(bp, (struct mbuf *)&mh, direction);
   1045 }
   1046 
   1047 /*
   1048  * Convert Ethernet address to printable (loggable) representation.
   1049  */
   1050 char *
   1051 ether_sprintf(const u_char *ap)
   1052 {
   1053 	static char etherbuf[3 * ETHER_ADDR_LEN];
   1054 	return ether_snprintf(etherbuf, sizeof(etherbuf), ap);
   1055 }
   1056 
   1057 char *
   1058 ether_snprintf(char *buf, size_t len, const u_char *ap)
   1059 {
   1060 	char *cp = buf;
   1061 	size_t i;
   1062 
   1063 	for (i = 0; i < len / 3; i++) {
   1064 		*cp++ = hexdigits[*ap >> 4];
   1065 		*cp++ = hexdigits[*ap++ & 0xf];
   1066 		*cp++ = ':';
   1067 	}
   1068 	*--cp = '\0';
   1069 	return buf;
   1070 }
   1071 
   1072 /*
   1073  * Perform common duties while attaching to interface list
   1074  */
   1075 void
   1076 ether_ifattach(struct ifnet *ifp, const uint8_t *lla)
   1077 {
   1078 	struct ethercom *ec = (struct ethercom *)ifp;
   1079 	char xnamebuf[HOOKNAMSIZ];
   1080 
   1081 	if (lla != NULL && ETHER_IS_MULTICAST(lla))
   1082 		aprint_error("The multicast bit is set in the MAC address. "
   1083 			"It's wrong.\n");
   1084 
   1085 	ifp->if_type = IFT_ETHER;
   1086 	ifp->if_hdrlen = ETHER_HDR_LEN;
   1087 	ifp->if_dlt = DLT_EN10MB;
   1088 	ifp->if_mtu = ETHERMTU;
   1089 	ifp->if_output = ether_output;
   1090 	ifp->_if_input = ether_input;
   1091 	if (ec->ec_capabilities & ETHERCAP_VLAN_HWTAGGING)
   1092 		ifp->if_bpf_mtap = ether_bpf_mtap;
   1093 	if (ifp->if_baudrate == 0)
   1094 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
   1095 
   1096 	if (lla != NULL)
   1097 		if_set_sadl(ifp, lla, ETHER_ADDR_LEN, !ETHER_IS_LOCAL(lla));
   1098 
   1099 	LIST_INIT(&ec->ec_multiaddrs);
   1100 	SIMPLEQ_INIT(&ec->ec_vids);
   1101 	ec->ec_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NET);
   1102 	ec->ec_flags = 0;
   1103 	ifp->if_broadcastaddr = etherbroadcastaddr;
   1104 	bpf_attach(ifp, DLT_EN10MB, sizeof(struct ether_header));
   1105 	snprintf(xnamebuf, sizeof(xnamebuf),
   1106 	    "%s-ether_ifdetachhooks", ifp->if_xname);
   1107 	ec->ec_ifdetach_hooks = simplehook_create(IPL_NET, xnamebuf);
   1108 #ifdef MBUFTRACE
   1109 	mowner_init_owner(&ec->ec_tx_mowner, ifp->if_xname, "tx");
   1110 	mowner_init_owner(&ec->ec_rx_mowner, ifp->if_xname, "rx");
   1111 	MOWNER_ATTACH(&ec->ec_tx_mowner);
   1112 	MOWNER_ATTACH(&ec->ec_rx_mowner);
   1113 	ifp->if_mowner = &ec->ec_tx_mowner;
   1114 #endif
   1115 }
   1116 
   1117 void
   1118 ether_ifdetach(struct ifnet *ifp)
   1119 {
   1120 	struct ethercom *ec = (void *) ifp;
   1121 	struct ether_multi *enm;
   1122 
   1123 	IFNET_ASSERT_UNLOCKED(ifp);
   1124 	/*
   1125 	 * Prevent further calls to ioctl (for example turning off
   1126 	 * promiscuous mode from the bridge code), which eventually can
   1127 	 * call if_init() which can cause panics because the interface
   1128 	 * is in the process of being detached. Return device not configured
   1129 	 * instead.
   1130 	 */
   1131 	ifp->if_ioctl = __FPTRCAST(int (*)(struct ifnet *, u_long, void *),
   1132 	    enxio);
   1133 
   1134 	simplehook_dohooks(ec->ec_ifdetach_hooks);
   1135 	KASSERT(!simplehook_has_hooks(ec->ec_ifdetach_hooks));
   1136 	simplehook_destroy(ec->ec_ifdetach_hooks);
   1137 
   1138 	bpf_detach(ifp);
   1139 
   1140 	ETHER_LOCK(ec);
   1141 	KASSERT(ec->ec_nvlans == 0);
   1142 	while ((enm = LIST_FIRST(&ec->ec_multiaddrs)) != NULL) {
   1143 		LIST_REMOVE(enm, enm_list);
   1144 		kmem_free(enm, sizeof(*enm));
   1145 		ec->ec_multicnt--;
   1146 	}
   1147 	ETHER_UNLOCK(ec);
   1148 
   1149 	mutex_obj_free(ec->ec_lock);
   1150 	ec->ec_lock = NULL;
   1151 
   1152 	ifp->if_mowner = NULL;
   1153 	MOWNER_DETACH(&ec->ec_rx_mowner);
   1154 	MOWNER_DETACH(&ec->ec_tx_mowner);
   1155 }
   1156 
   1157 void *
   1158 ether_ifdetachhook_establish(struct ifnet *ifp,
   1159     void (*fn)(void *), void *arg)
   1160 {
   1161 	struct ethercom *ec;
   1162 	khook_t *hk;
   1163 
   1164 	if (ifp->if_type != IFT_ETHER)
   1165 		return NULL;
   1166 
   1167 	ec = (struct ethercom *)ifp;
   1168 	hk = simplehook_establish(ec->ec_ifdetach_hooks,
   1169 	    fn, arg);
   1170 
   1171 	return (void *)hk;
   1172 }
   1173 
   1174 void
   1175 ether_ifdetachhook_disestablish(struct ifnet *ifp,
   1176     void *vhook, kmutex_t *lock)
   1177 {
   1178 	struct ethercom *ec;
   1179 
   1180 	if (vhook == NULL)
   1181 		return;
   1182 
   1183 	ec = (struct ethercom *)ifp;
   1184 	simplehook_disestablish(ec->ec_ifdetach_hooks, vhook, lock);
   1185 }
   1186 
   1187 #if 0
   1188 /*
   1189  * This is for reference.  We have a table-driven version
   1190  * of the little-endian crc32 generator, which is faster
   1191  * than the double-loop.
   1192  */
   1193 uint32_t
   1194 ether_crc32_le(const uint8_t *buf, size_t len)
   1195 {
   1196 	uint32_t c, crc, carry;
   1197 	size_t i, j;
   1198 
   1199 	crc = 0xffffffffU;	/* initial value */
   1200 
   1201 	for (i = 0; i < len; i++) {
   1202 		c = buf[i];
   1203 		for (j = 0; j < 8; j++) {
   1204 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
   1205 			crc >>= 1;
   1206 			c >>= 1;
   1207 			if (carry)
   1208 				crc = (crc ^ ETHER_CRC_POLY_LE);
   1209 		}
   1210 	}
   1211 
   1212 	return (crc);
   1213 }
   1214 #else
   1215 uint32_t
   1216 ether_crc32_le(const uint8_t *buf, size_t len)
   1217 {
   1218 	static const uint32_t crctab[] = {
   1219 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
   1220 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
   1221 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
   1222 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
   1223 	};
   1224 	uint32_t crc;
   1225 	size_t i;
   1226 
   1227 	crc = 0xffffffffU;	/* initial value */
   1228 
   1229 	for (i = 0; i < len; i++) {
   1230 		crc ^= buf[i];
   1231 		crc = (crc >> 4) ^ crctab[crc & 0xf];
   1232 		crc = (crc >> 4) ^ crctab[crc & 0xf];
   1233 	}
   1234 
   1235 	return (crc);
   1236 }
   1237 #endif
   1238 
   1239 uint32_t
   1240 ether_crc32_be(const uint8_t *buf, size_t len)
   1241 {
   1242 	uint32_t c, crc, carry;
   1243 	size_t i, j;
   1244 
   1245 	crc = 0xffffffffU;	/* initial value */
   1246 
   1247 	for (i = 0; i < len; i++) {
   1248 		c = buf[i];
   1249 		for (j = 0; j < 8; j++) {
   1250 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
   1251 			crc <<= 1;
   1252 			c >>= 1;
   1253 			if (carry)
   1254 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
   1255 		}
   1256 	}
   1257 
   1258 	return (crc);
   1259 }
   1260 
   1261 #ifdef INET
   1262 const uint8_t ether_ipmulticast_min[ETHER_ADDR_LEN] =
   1263     { 0x01, 0x00, 0x5e, 0x00, 0x00, 0x00 };
   1264 const uint8_t ether_ipmulticast_max[ETHER_ADDR_LEN] =
   1265     { 0x01, 0x00, 0x5e, 0x7f, 0xff, 0xff };
   1266 #endif
   1267 #ifdef INET6
   1268 const uint8_t ether_ip6multicast_min[ETHER_ADDR_LEN] =
   1269     { 0x33, 0x33, 0x00, 0x00, 0x00, 0x00 };
   1270 const uint8_t ether_ip6multicast_max[ETHER_ADDR_LEN] =
   1271     { 0x33, 0x33, 0xff, 0xff, 0xff, 0xff };
   1272 #endif
   1273 
   1274 /*
   1275  * ether_aton implementation, not using a static buffer.
   1276  */
   1277 int
   1278 ether_aton_r(u_char *dest, size_t len, const char *str)
   1279 {
   1280 	const u_char *cp = (const void *)str;
   1281 	u_char *ep;
   1282 
   1283 #define atox(c)	(((c) <= '9') ? ((c) - '0') : ((toupper(c) - 'A') + 10))
   1284 
   1285 	if (len < ETHER_ADDR_LEN)
   1286 		return ENOSPC;
   1287 
   1288 	ep = dest + ETHER_ADDR_LEN;
   1289 
   1290 	while (*cp) {
   1291 		if (!isxdigit(*cp))
   1292 			return EINVAL;
   1293 
   1294 		*dest = atox(*cp);
   1295 		cp++;
   1296 		if (isxdigit(*cp)) {
   1297 			*dest = (*dest << 4) | atox(*cp);
   1298 			cp++;
   1299 		}
   1300 		dest++;
   1301 
   1302 		if (dest == ep)
   1303 			return (*cp == '\0') ? 0 : ENAMETOOLONG;
   1304 
   1305 		switch (*cp) {
   1306 		case ':':
   1307 		case '-':
   1308 		case '.':
   1309 			cp++;
   1310 			break;
   1311 		}
   1312 	}
   1313 	return ENOBUFS;
   1314 }
   1315 
   1316 /*
   1317  * Convert a sockaddr into an Ethernet address or range of Ethernet
   1318  * addresses.
   1319  */
   1320 int
   1321 ether_multiaddr(const struct sockaddr *sa, uint8_t addrlo[ETHER_ADDR_LEN],
   1322     uint8_t addrhi[ETHER_ADDR_LEN])
   1323 {
   1324 #ifdef INET
   1325 	const struct sockaddr_in *sin;
   1326 #endif
   1327 #ifdef INET6
   1328 	const struct sockaddr_in6 *sin6;
   1329 #endif
   1330 
   1331 	switch (sa->sa_family) {
   1332 
   1333 	case AF_UNSPEC:
   1334 		memcpy(addrlo, sa->sa_data, ETHER_ADDR_LEN);
   1335 		memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
   1336 		break;
   1337 
   1338 #ifdef INET
   1339 	case AF_INET:
   1340 		sin = satocsin(sa);
   1341 		if (sin->sin_addr.s_addr == INADDR_ANY) {
   1342 			/*
   1343 			 * An IP address of INADDR_ANY means listen to
   1344 			 * or stop listening to all of the Ethernet
   1345 			 * multicast addresses used for IP.
   1346 			 * (This is for the sake of IP multicast routers.)
   1347 			 */
   1348 			memcpy(addrlo, ether_ipmulticast_min, ETHER_ADDR_LEN);
   1349 			memcpy(addrhi, ether_ipmulticast_max, ETHER_ADDR_LEN);
   1350 		} else {
   1351 			ETHER_MAP_IP_MULTICAST(&sin->sin_addr, addrlo);
   1352 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
   1353 		}
   1354 		break;
   1355 #endif
   1356 #ifdef INET6
   1357 	case AF_INET6:
   1358 		sin6 = satocsin6(sa);
   1359 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
   1360 			/*
   1361 			 * An IP6 address of 0 means listen to or stop
   1362 			 * listening to all of the Ethernet multicast
   1363 			 * address used for IP6.
   1364 			 * (This is used for multicast routers.)
   1365 			 */
   1366 			memcpy(addrlo, ether_ip6multicast_min, ETHER_ADDR_LEN);
   1367 			memcpy(addrhi, ether_ip6multicast_max, ETHER_ADDR_LEN);
   1368 		} else {
   1369 			ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, addrlo);
   1370 			memcpy(addrhi, addrlo, ETHER_ADDR_LEN);
   1371 		}
   1372 		break;
   1373 #endif
   1374 
   1375 	default:
   1376 		return EAFNOSUPPORT;
   1377 	}
   1378 	return 0;
   1379 }
   1380 
   1381 /*
   1382  * Add an Ethernet multicast address or range of addresses to the list for a
   1383  * given interface.
   1384  */
   1385 int
   1386 ether_addmulti(const struct sockaddr *sa, struct ethercom *ec)
   1387 {
   1388 	struct ether_multi *enm, *_enm;
   1389 	u_char addrlo[ETHER_ADDR_LEN];
   1390 	u_char addrhi[ETHER_ADDR_LEN];
   1391 	int error = 0;
   1392 
   1393 	/* Allocate out of lock */
   1394 	enm = kmem_alloc(sizeof(*enm), KM_SLEEP);
   1395 
   1396 	ETHER_LOCK(ec);
   1397 	error = ether_multiaddr(sa, addrlo, addrhi);
   1398 	if (error != 0)
   1399 		goto out;
   1400 
   1401 	/*
   1402 	 * Verify that we have valid Ethernet multicast addresses.
   1403 	 */
   1404 	if (!ETHER_IS_MULTICAST(addrlo) || !ETHER_IS_MULTICAST(addrhi)) {
   1405 		error = EINVAL;
   1406 		goto out;
   1407 	}
   1408 
   1409 	/*
   1410 	 * See if the address range is already in the list.
   1411 	 */
   1412 	_enm = ether_lookup_multi(addrlo, addrhi, ec);
   1413 	if (_enm != NULL) {
   1414 		/*
   1415 		 * Found it; just increment the reference count.
   1416 		 */
   1417 		++_enm->enm_refcount;
   1418 		error = 0;
   1419 		goto out;
   1420 	}
   1421 
   1422 	/*
   1423 	 * Link a new multicast record into the interface's multicast list.
   1424 	 */
   1425 	memcpy(enm->enm_addrlo, addrlo, ETHER_ADDR_LEN);
   1426 	memcpy(enm->enm_addrhi, addrhi, ETHER_ADDR_LEN);
   1427 	enm->enm_refcount = 1;
   1428 	LIST_INSERT_HEAD(&ec->ec_multiaddrs, enm, enm_list);
   1429 	ec->ec_multicnt++;
   1430 
   1431 	/*
   1432 	 * Return ENETRESET to inform the driver that the list has changed
   1433 	 * and its reception filter should be adjusted accordingly.
   1434 	 */
   1435 	error = ENETRESET;
   1436 	enm = NULL;
   1437 
   1438 out:
   1439 	ETHER_UNLOCK(ec);
   1440 	if (enm != NULL)
   1441 		kmem_free(enm, sizeof(*enm));
   1442 	return error;
   1443 }
   1444 
   1445 /*
   1446  * Delete a multicast address record.
   1447  */
   1448 int
   1449 ether_delmulti(const struct sockaddr *sa, struct ethercom *ec)
   1450 {
   1451 	struct ether_multi *enm;
   1452 	u_char addrlo[ETHER_ADDR_LEN];
   1453 	u_char addrhi[ETHER_ADDR_LEN];
   1454 	int error;
   1455 
   1456 	ETHER_LOCK(ec);
   1457 	error = ether_multiaddr(sa, addrlo, addrhi);
   1458 	if (error != 0)
   1459 		goto error;
   1460 
   1461 	/*
   1462 	 * Look up the address in our list.
   1463 	 */
   1464 	enm = ether_lookup_multi(addrlo, addrhi, ec);
   1465 	if (enm == NULL) {
   1466 		error = ENXIO;
   1467 		goto error;
   1468 	}
   1469 	if (--enm->enm_refcount != 0) {
   1470 		/*
   1471 		 * Still some claims to this record.
   1472 		 */
   1473 		error = 0;
   1474 		goto error;
   1475 	}
   1476 
   1477 	/*
   1478 	 * No remaining claims to this record; unlink and free it.
   1479 	 */
   1480 	LIST_REMOVE(enm, enm_list);
   1481 	ec->ec_multicnt--;
   1482 	ETHER_UNLOCK(ec);
   1483 	kmem_free(enm, sizeof(*enm));
   1484 
   1485 	/*
   1486 	 * Return ENETRESET to inform the driver that the list has changed
   1487 	 * and its reception filter should be adjusted accordingly.
   1488 	 */
   1489 	return ENETRESET;
   1490 
   1491 error:
   1492 	ETHER_UNLOCK(ec);
   1493 	return error;
   1494 }
   1495 
   1496 void
   1497 ether_set_ifflags_cb(struct ethercom *ec, ether_cb_t cb)
   1498 {
   1499 	ec->ec_ifflags_cb = cb;
   1500 }
   1501 
   1502 void
   1503 ether_set_vlan_cb(struct ethercom *ec, ether_vlancb_t cb)
   1504 {
   1505 
   1506 	ec->ec_vlan_cb = cb;
   1507 }
   1508 
   1509 static int
   1510 ether_ioctl_reinit(struct ethercom *ec)
   1511 {
   1512 	struct ifnet *ifp = &ec->ec_if;
   1513 	int error;
   1514 
   1515 	KASSERTMSG(IFNET_LOCKED(ifp), "%s", ifp->if_xname);
   1516 
   1517 	switch (ifp->if_flags & (IFF_UP | IFF_RUNNING)) {
   1518 	case IFF_RUNNING:
   1519 		/*
   1520 		 * If interface is marked down and it is running,
   1521 		 * then stop and disable it.
   1522 		 */
   1523 		if_stop(ifp, 1);
   1524 		break;
   1525 	case IFF_UP:
   1526 		/*
   1527 		 * If interface is marked up and it is stopped, then
   1528 		 * start it.
   1529 		 */
   1530 		return if_init(ifp);
   1531 	case IFF_UP | IFF_RUNNING:
   1532 		error = 0;
   1533 		if (ec->ec_ifflags_cb != NULL) {
   1534 			error = (*ec->ec_ifflags_cb)(ec);
   1535 			if (error == ENETRESET) {
   1536 				/*
   1537 				 * Reset the interface to pick up
   1538 				 * changes in any other flags that
   1539 				 * affect the hardware state.
   1540 				 */
   1541 				return if_init(ifp);
   1542 			}
   1543 		} else
   1544 			error = if_init(ifp);
   1545 		return error;
   1546 	case 0:
   1547 		break;
   1548 	}
   1549 
   1550 	return 0;
   1551 }
   1552 
   1553 /*
   1554  * Common ioctls for Ethernet interfaces.  Note, we must be
   1555  * called at splnet().
   1556  */
   1557 int
   1558 ether_ioctl(struct ifnet *ifp, u_long cmd, void *data)
   1559 {
   1560 	struct ethercom *ec = (void *)ifp;
   1561 	struct eccapreq *eccr;
   1562 	struct ifreq *ifr = (struct ifreq *)data;
   1563 	struct if_laddrreq *iflr = data;
   1564 	const struct sockaddr_dl *sdl;
   1565 	static const uint8_t zero[ETHER_ADDR_LEN];
   1566 	int error;
   1567 
   1568 	switch (cmd) {
   1569 	case SIOCINITIFADDR:
   1570 	    {
   1571 		struct ifaddr *ifa = (struct ifaddr *)data;
   1572 		if (ifa->ifa_addr->sa_family != AF_LINK
   1573 		    && (ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
   1574 		       (IFF_UP | IFF_RUNNING)) {
   1575 			ifp->if_flags |= IFF_UP;
   1576 			if ((error = if_init(ifp)) != 0)
   1577 				return error;
   1578 		}
   1579 #ifdef INET
   1580 		if (ifa->ifa_addr->sa_family == AF_INET)
   1581 			arp_ifinit(ifp, ifa);
   1582 #endif
   1583 		return 0;
   1584 	    }
   1585 
   1586 	case SIOCSIFMTU:
   1587 	    {
   1588 		int maxmtu;
   1589 
   1590 		if (ec->ec_capabilities & ETHERCAP_JUMBO_MTU)
   1591 			maxmtu = ETHERMTU_JUMBO;
   1592 		else
   1593 			maxmtu = ETHERMTU;
   1594 
   1595 		if (ifr->ifr_mtu < ETHERMIN || ifr->ifr_mtu > maxmtu)
   1596 			return EINVAL;
   1597 		else if ((error = ifioctl_common(ifp, cmd, data)) != ENETRESET)
   1598 			return error;
   1599 		else if (ifp->if_flags & IFF_UP) {
   1600 			/* Make sure the device notices the MTU change. */
   1601 			return if_init(ifp);
   1602 		} else
   1603 			return 0;
   1604 	    }
   1605 
   1606 	case SIOCSIFFLAGS:
   1607 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
   1608 			return error;
   1609 		return ether_ioctl_reinit(ec);
   1610 	case SIOCGIFFLAGS:
   1611 		error = ifioctl_common(ifp, cmd, data);
   1612 		if (error == 0) {
   1613 			/* Set IFF_ALLMULTI for backcompat */
   1614 			ifr->ifr_flags |= (ec->ec_flags & ETHER_F_ALLMULTI) ?
   1615 			    IFF_ALLMULTI : 0;
   1616 		}
   1617 		return error;
   1618 	case SIOCGETHERCAP:
   1619 		eccr = (struct eccapreq *)data;
   1620 		eccr->eccr_capabilities = ec->ec_capabilities;
   1621 		eccr->eccr_capenable = ec->ec_capenable;
   1622 		return 0;
   1623 	case SIOCSETHERCAP:
   1624 		eccr = (struct eccapreq *)data;
   1625 		if ((eccr->eccr_capenable & ~ec->ec_capabilities) != 0)
   1626 			return EINVAL;
   1627 		if (eccr->eccr_capenable == ec->ec_capenable)
   1628 			return 0;
   1629 #if 0 /* notyet */
   1630 		ec->ec_capenable = (ec->ec_capenable & ETHERCAP_CANTCHANGE)
   1631 		    | (eccr->eccr_capenable & ~ETHERCAP_CANTCHANGE);
   1632 #else
   1633 		ec->ec_capenable = eccr->eccr_capenable;
   1634 #endif
   1635 		return ether_ioctl_reinit(ec);
   1636 	case SIOCADDMULTI:
   1637 		return ether_addmulti(ifreq_getaddr(cmd, ifr), ec);
   1638 	case SIOCDELMULTI:
   1639 		return ether_delmulti(ifreq_getaddr(cmd, ifr), ec);
   1640 	case SIOCSIFMEDIA:
   1641 	case SIOCGIFMEDIA:
   1642 		if (ec->ec_mii != NULL)
   1643 			return ifmedia_ioctl(ifp, ifr, &ec->ec_mii->mii_media,
   1644 			    cmd);
   1645 		else if (ec->ec_ifmedia != NULL)
   1646 			return ifmedia_ioctl(ifp, ifr, ec->ec_ifmedia, cmd);
   1647 		else
   1648 			return ENOTTY;
   1649 		break;
   1650 	case SIOCALIFADDR:
   1651 		sdl = satocsdl(sstocsa(&iflr->addr));
   1652 		if (sdl->sdl_family != AF_LINK)
   1653 			;
   1654 		else if (ETHER_IS_MULTICAST(CLLADDR(sdl)))
   1655 			return EINVAL;
   1656 		else if (memcmp(zero, CLLADDR(sdl), sizeof(zero)) == 0)
   1657 			return EINVAL;
   1658 		/*FALLTHROUGH*/
   1659 	default:
   1660 		return ifioctl_common(ifp, cmd, data);
   1661 	}
   1662 	return 0;
   1663 }
   1664 
   1665 /*
   1666  * Enable/disable passing VLAN packets if the parent interface supports it.
   1667  * Return:
   1668  * 	 0: Ok
   1669  *	-1: Parent interface does not support vlans
   1670  *	>0: Error
   1671  */
   1672 int
   1673 ether_enable_vlan_mtu(struct ifnet *ifp)
   1674 {
   1675 	int error;
   1676 	struct ethercom *ec = (void *)ifp;
   1677 
   1678 	/* Parent does not support VLAN's */
   1679 	if ((ec->ec_capabilities & ETHERCAP_VLAN_MTU) == 0)
   1680 		return -1;
   1681 
   1682 	/*
   1683 	 * Parent supports the VLAN_MTU capability,
   1684 	 * i.e. can Tx/Rx larger than ETHER_MAX_LEN frames;
   1685 	 * enable it.
   1686 	 */
   1687 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
   1688 
   1689 	/* Interface is down, defer for later */
   1690 	if ((ifp->if_flags & IFF_UP) == 0)
   1691 		return 0;
   1692 
   1693 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
   1694 		return 0;
   1695 
   1696 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
   1697 	return error;
   1698 }
   1699 
   1700 int
   1701 ether_disable_vlan_mtu(struct ifnet *ifp)
   1702 {
   1703 	int error;
   1704 	struct ethercom *ec = (void *)ifp;
   1705 
   1706 	/* We still have VLAN's, defer for later */
   1707 	if (ec->ec_nvlans != 0)
   1708 		return 0;
   1709 
   1710 	/* Parent does not support VLAB's, nothing to do. */
   1711 	if ((ec->ec_capenable & ETHERCAP_VLAN_MTU) == 0)
   1712 		return -1;
   1713 
   1714 	/*
   1715 	 * Disable Tx/Rx of VLAN-sized frames.
   1716 	 */
   1717 	ec->ec_capenable &= ~ETHERCAP_VLAN_MTU;
   1718 
   1719 	/* Interface is down, defer for later */
   1720 	if ((ifp->if_flags & IFF_UP) == 0)
   1721 		return 0;
   1722 
   1723 	if ((error = if_flags_set(ifp, ifp->if_flags)) == 0)
   1724 		return 0;
   1725 
   1726 	ec->ec_capenable |= ETHERCAP_VLAN_MTU;
   1727 	return error;
   1728 }
   1729 
   1730 /*
   1731  * Add and delete VLAN TAG
   1732  */
   1733 int
   1734 ether_add_vlantag(struct ifnet *ifp, uint16_t vtag, bool *vlanmtu_status)
   1735 {
   1736 	struct ethercom *ec = (void *)ifp;
   1737 	struct vlanid_list *vidp;
   1738 	bool vlanmtu_enabled;
   1739 	uint16_t vid = EVL_VLANOFTAG(vtag);
   1740 	int error;
   1741 
   1742 	vlanmtu_enabled = false;
   1743 
   1744 	/* Add a vid to the list */
   1745 	vidp = kmem_alloc(sizeof(*vidp), KM_SLEEP);
   1746 	vidp->vid = vid;
   1747 
   1748 	ETHER_LOCK(ec);
   1749 	ec->ec_nvlans++;
   1750 	SIMPLEQ_INSERT_TAIL(&ec->ec_vids, vidp, vid_list);
   1751 	ETHER_UNLOCK(ec);
   1752 
   1753 	if (ec->ec_nvlans == 1) {
   1754 		IFNET_LOCK(ifp);
   1755 		error = ether_enable_vlan_mtu(ifp);
   1756 		IFNET_UNLOCK(ifp);
   1757 
   1758 		if (error == 0) {
   1759 			vlanmtu_enabled = true;
   1760 		} else if (error != -1) {
   1761 			goto fail;
   1762 		}
   1763 	}
   1764 
   1765 	if (ec->ec_vlan_cb != NULL) {
   1766 		error = (*ec->ec_vlan_cb)(ec, vid, true);
   1767 		if (error != 0)
   1768 			goto fail;
   1769 	}
   1770 
   1771 	if (vlanmtu_status != NULL)
   1772 		*vlanmtu_status = vlanmtu_enabled;
   1773 
   1774 	return 0;
   1775 fail:
   1776 	ETHER_LOCK(ec);
   1777 	ec->ec_nvlans--;
   1778 	SIMPLEQ_REMOVE(&ec->ec_vids, vidp, vlanid_list, vid_list);
   1779 	ETHER_UNLOCK(ec);
   1780 
   1781 	if (vlanmtu_enabled) {
   1782 		IFNET_LOCK(ifp);
   1783 		(void)ether_disable_vlan_mtu(ifp);
   1784 		IFNET_UNLOCK(ifp);
   1785 	}
   1786 
   1787 	kmem_free(vidp, sizeof(*vidp));
   1788 
   1789 	return error;
   1790 }
   1791 
   1792 int
   1793 ether_del_vlantag(struct ifnet *ifp, uint16_t vtag)
   1794 {
   1795 	struct ethercom *ec = (void *)ifp;
   1796 	struct vlanid_list *vidp;
   1797 	uint16_t vid = EVL_VLANOFTAG(vtag);
   1798 
   1799 	ETHER_LOCK(ec);
   1800 	SIMPLEQ_FOREACH(vidp, &ec->ec_vids, vid_list) {
   1801 		if (vidp->vid == vid) {
   1802 			SIMPLEQ_REMOVE(&ec->ec_vids, vidp,
   1803 			    vlanid_list, vid_list);
   1804 			ec->ec_nvlans--;
   1805 			break;
   1806 		}
   1807 	}
   1808 	ETHER_UNLOCK(ec);
   1809 
   1810 	if (vidp == NULL)
   1811 		return ENOENT;
   1812 
   1813 	if (ec->ec_vlan_cb != NULL) {
   1814 		(void)(*ec->ec_vlan_cb)(ec, vidp->vid, false);
   1815 	}
   1816 
   1817 	if (ec->ec_nvlans == 0) {
   1818 		IFNET_LOCK(ifp);
   1819 		(void)ether_disable_vlan_mtu(ifp);
   1820 		IFNET_UNLOCK(ifp);
   1821 	}
   1822 
   1823 	kmem_free(vidp, sizeof(*vidp));
   1824 
   1825 	return 0;
   1826 }
   1827 
   1828 int
   1829 ether_inject_vlantag(struct mbuf **mp, uint16_t etype, uint16_t tag)
   1830 {
   1831 	static const size_t min_data_len =
   1832 	    ETHER_MIN_LEN - ETHER_CRC_LEN + ETHER_VLAN_ENCAP_LEN;
   1833 	/* Used to pad ethernet frames with < ETHER_MIN_LEN bytes */
   1834 	static const char vlan_zero_pad_buff[ETHER_MIN_LEN] = { 0 };
   1835 
   1836 	struct ether_vlan_header *evl;
   1837 	struct mbuf *m = *mp;
   1838 	int error;
   1839 
   1840 	error = 0;
   1841 
   1842 	M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_DONTWAIT);
   1843 	if (m == NULL) {
   1844 		error = ENOBUFS;
   1845 		goto out;
   1846 	}
   1847 
   1848 	if (m->m_len < sizeof(*evl)) {
   1849 		m = m_pullup(m, sizeof(*evl));
   1850 		if (m == NULL) {
   1851 			error = ENOBUFS;
   1852 			goto out;
   1853 		}
   1854 	}
   1855 
   1856 	/*
   1857 	 * Transform the Ethernet header into an
   1858 	 * Ethernet header with 802.1Q encapsulation.
   1859 	 */
   1860 	memmove(mtod(m, void *),
   1861 	    mtod(m, char *) + ETHER_VLAN_ENCAP_LEN,
   1862 	    sizeof(struct ether_header));
   1863 	evl = mtod(m, struct ether_vlan_header *);
   1864 	evl->evl_proto = evl->evl_encap_proto;
   1865 	evl->evl_encap_proto = htons(etype);
   1866 	evl->evl_tag = htons(tag);
   1867 
   1868 	/*
   1869 	 * To cater for VLAN-aware layer 2 ethernet
   1870 	 * switches which may need to strip the tag
   1871 	 * before forwarding the packet, make sure
   1872 	 * the packet+tag is at least 68 bytes long.
   1873 	 * This is necessary because our parent will
   1874 	 * only pad to 64 bytes (ETHER_MIN_LEN) and
   1875 	 * some switches will not pad by themselves
   1876 	 * after deleting a tag.
   1877 	 */
   1878 	if (m->m_pkthdr.len < min_data_len) {
   1879 		m_copyback(m, m->m_pkthdr.len,
   1880 		    min_data_len - m->m_pkthdr.len,
   1881 		    vlan_zero_pad_buff);
   1882 	}
   1883 
   1884 	m->m_flags &= ~M_VLANTAG;
   1885 
   1886 out:
   1887 	*mp = m;
   1888 	return error;
   1889 }
   1890 
   1891 struct mbuf *
   1892 ether_strip_vlantag(struct mbuf *m)
   1893 {
   1894 	struct ether_vlan_header *evl;
   1895 
   1896 	if (m->m_len < sizeof(*evl) &&
   1897 	    (m = m_pullup(m, sizeof(*evl))) == NULL) {
   1898 		return NULL;
   1899 	}
   1900 
   1901 	if (m_makewritable(&m, 0, sizeof(*evl), M_DONTWAIT)) {
   1902 		m_freem(m);
   1903 		return NULL;
   1904 	}
   1905 
   1906 	evl = mtod(m, struct ether_vlan_header *);
   1907 	KASSERT(ntohs(evl->evl_encap_proto) == ETHERTYPE_VLAN);
   1908 
   1909 	vlan_set_tag(m, ntohs(evl->evl_tag));
   1910 
   1911 	/*
   1912 	 * Restore the original ethertype.  We'll remove
   1913 	 * the encapsulation after we've found the vlan
   1914 	 * interface corresponding to the tag.
   1915 	 */
   1916 	evl->evl_encap_proto = evl->evl_proto;
   1917 
   1918 	/*
   1919 	 * Remove the encapsulation header and append tag.
   1920 	 * The original header has already been fixed up above.
   1921 	 */
   1922 	vlan_set_tag(m, ntohs(evl->evl_tag));
   1923 	memmove((char *)evl + ETHER_VLAN_ENCAP_LEN, evl,
   1924 	    offsetof(struct ether_vlan_header, evl_encap_proto));
   1925 	m_adj(m, ETHER_VLAN_ENCAP_LEN);
   1926 
   1927 	return m;
   1928 }
   1929 
   1930 static int
   1931 ether_multicast_sysctl(SYSCTLFN_ARGS)
   1932 {
   1933 	struct ether_multi *enm;
   1934 	struct ifnet *ifp;
   1935 	struct ethercom *ec;
   1936 	int error = 0;
   1937 	size_t written;
   1938 	struct psref psref;
   1939 	int bound;
   1940 	unsigned int multicnt;
   1941 	struct ether_multi_sysctl *addrs;
   1942 	int i;
   1943 
   1944 	if (namelen != 1)
   1945 		return EINVAL;
   1946 
   1947 	bound = curlwp_bind();
   1948 	ifp = if_get_byindex(name[0], &psref);
   1949 	if (ifp == NULL) {
   1950 		error = ENODEV;
   1951 		goto out;
   1952 	}
   1953 	if (ifp->if_type != IFT_ETHER) {
   1954 		if_put(ifp, &psref);
   1955 		*oldlenp = 0;
   1956 		goto out;
   1957 	}
   1958 	ec = (struct ethercom *)ifp;
   1959 
   1960 	if (oldp == NULL) {
   1961 		if_put(ifp, &psref);
   1962 		*oldlenp = ec->ec_multicnt * sizeof(*addrs);
   1963 		goto out;
   1964 	}
   1965 
   1966 	/*
   1967 	 * ec->ec_lock is a spin mutex so we cannot call sysctl_copyout, which
   1968 	 * is sleepable, while holding it. Copy data to a local buffer first
   1969 	 * with the lock taken and then call sysctl_copyout without holding it.
   1970 	 */
   1971 retry:
   1972 	multicnt = ec->ec_multicnt;
   1973 
   1974 	if (multicnt == 0) {
   1975 		if_put(ifp, &psref);
   1976 		*oldlenp = 0;
   1977 		goto out;
   1978 	}
   1979 
   1980 	addrs = kmem_zalloc(sizeof(*addrs) * multicnt, KM_SLEEP);
   1981 
   1982 	ETHER_LOCK(ec);
   1983 	if (multicnt != ec->ec_multicnt) {
   1984 		/* The number of multicast addresses has changed */
   1985 		ETHER_UNLOCK(ec);
   1986 		kmem_free(addrs, sizeof(*addrs) * multicnt);
   1987 		goto retry;
   1988 	}
   1989 
   1990 	i = 0;
   1991 	LIST_FOREACH(enm, &ec->ec_multiaddrs, enm_list) {
   1992 		struct ether_multi_sysctl *addr = &addrs[i];
   1993 		addr->enm_refcount = enm->enm_refcount;
   1994 		memcpy(addr->enm_addrlo, enm->enm_addrlo, ETHER_ADDR_LEN);
   1995 		memcpy(addr->enm_addrhi, enm->enm_addrhi, ETHER_ADDR_LEN);
   1996 		i++;
   1997 	}
   1998 	ETHER_UNLOCK(ec);
   1999 
   2000 	error = 0;
   2001 	written = 0;
   2002 	for (i = 0; i < multicnt; i++) {
   2003 		struct ether_multi_sysctl *addr = &addrs[i];
   2004 
   2005 		if (written + sizeof(*addr) > *oldlenp)
   2006 			break;
   2007 		error = sysctl_copyout(l, addr, oldp, sizeof(*addr));
   2008 		if (error)
   2009 			break;
   2010 		written += sizeof(*addr);
   2011 		oldp = (char *)oldp + sizeof(*addr);
   2012 	}
   2013 	kmem_free(addrs, sizeof(*addrs) * multicnt);
   2014 
   2015 	if_put(ifp, &psref);
   2016 
   2017 	*oldlenp = written;
   2018 out:
   2019 	curlwp_bindx(bound);
   2020 	return error;
   2021 }
   2022 
   2023 static void
   2024 ether_sysctl_setup(struct sysctllog **clog)
   2025 {
   2026 	const struct sysctlnode *rnode = NULL;
   2027 
   2028 	sysctl_createv(clog, 0, NULL, &rnode,
   2029 		       CTLFLAG_PERMANENT,
   2030 		       CTLTYPE_NODE, "ether",
   2031 		       SYSCTL_DESCR("Ethernet-specific information"),
   2032 		       NULL, 0, NULL, 0,
   2033 		       CTL_NET, CTL_CREATE, CTL_EOL);
   2034 
   2035 	sysctl_createv(clog, 0, &rnode, NULL,
   2036 		       CTLFLAG_PERMANENT,
   2037 		       CTLTYPE_NODE, "multicast",
   2038 		       SYSCTL_DESCR("multicast addresses"),
   2039 		       ether_multicast_sysctl, 0, NULL, 0,
   2040 		       CTL_CREATE, CTL_EOL);
   2041 
   2042 	sysctl_createv(clog, 0, &rnode, NULL,
   2043 		       CTLFLAG_PERMANENT | CTLFLAG_READWRITE,
   2044 		       CTLTYPE_STRING, "rps_hash",
   2045 		       SYSCTL_DESCR("Interface rps hash function control"),
   2046 		       sysctl_pktq_rps_hash_handler, 0, (void *)&ether_pktq_rps_hash_p,
   2047 		       PKTQ_RPS_HASH_NAME_LEN,
   2048 		       CTL_CREATE, CTL_EOL);
   2049 }
   2050 
   2051 void
   2052 etherinit(void)
   2053 {
   2054 
   2055 #ifdef DIAGNOSTIC
   2056 	mutex_init(&bigpktpps_lock, MUTEX_DEFAULT, IPL_NET);
   2057 #endif
   2058 	ether_pktq_rps_hash_p = pktq_rps_hash_default;
   2059 	ether_sysctl_setup(NULL);
   2060 }
   2061