2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1989, 1993
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
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 University 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.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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
31 * @(#)if_ethersubr.c 8.1 (Berkeley) 6/10/93
36 #include "opt_inet6.h"
37 #include "opt_netgraph.h"
38 #include "opt_mbuf_profiling.h"
41 #include <sys/param.h>
42 #include <sys/systm.h>
44 #include <sys/eventhandler.h>
45 #include <sys/kernel.h>
47 #include <sys/malloc.h>
48 #include <sys/module.h>
51 #include <sys/random.h>
52 #include <sys/socket.h>
53 #include <sys/sockio.h>
54 #include <sys/sysctl.h>
58 #include <net/if_var.h>
59 #include <net/if_arp.h>
60 #include <net/netisr.h>
61 #include <net/route.h>
62 #include <net/if_llc.h>
63 #include <net/if_dl.h>
64 #include <net/if_types.h>
66 #include <net/ethernet.h>
67 #include <net/if_bridgevar.h>
68 #include <net/if_vlan_var.h>
69 #include <net/if_llatbl.h>
71 #include <net/rss_config.h>
74 #include <netpfil/pf/pf_mtag.h>
76 #if defined(INET) || defined(INET6)
77 #include <netinet/in.h>
78 #include <netinet/in_var.h>
79 #include <netinet/if_ether.h>
80 #include <netinet/ip_carp.h>
81 #include <netinet/ip_var.h>
84 #include <netinet6/nd6.h>
86 #include <security/mac/mac_framework.h>
89 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2);
90 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN);
93 VNET_DEFINE(struct pfil_head, link_pfil_hook); /* Packet filter hooks */
95 /* netgraph node hooks for ng_ether(4) */
96 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
97 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
98 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
99 void (*ng_ether_attach_p)(struct ifnet *ifp);
100 void (*ng_ether_detach_p)(struct ifnet *ifp);
102 void (*vlan_input_p)(struct ifnet *, struct mbuf *);
104 /* if_bridge(4) support */
105 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
107 /* if_lagg(4) support */
108 struct mbuf *(*lagg_input_p)(struct ifnet *, struct mbuf *);
110 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
111 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
113 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
116 static void ether_reassign(struct ifnet *, struct vnet *, char *);
118 static int ether_requestencap(struct ifnet *, struct if_encap_req *);
121 #define senderr(e) do { error = (e); goto bad;} while (0)
124 update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst)
128 if (src->m_pkthdr.csum_flags & CSUM_IP)
129 csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
130 if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
131 csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
132 if (src->m_pkthdr.csum_flags & CSUM_SCTP)
133 csum_flags |= CSUM_SCTP_VALID;
134 dst->m_pkthdr.csum_flags |= csum_flags;
135 if (csum_flags & CSUM_DATA_VALID)
136 dst->m_pkthdr.csum_data = 0xffff;
140 * Handle link-layer encapsulation requests.
143 ether_requestencap(struct ifnet *ifp, struct if_encap_req *req)
145 struct ether_header *eh;
148 const u_char *lladdr;
150 if (req->rtype != IFENCAP_LL)
153 if (req->bufsize < ETHER_HDR_LEN)
156 eh = (struct ether_header *)req->buf;
157 lladdr = req->lladdr;
160 switch (req->family) {
162 etype = htons(ETHERTYPE_IP);
165 etype = htons(ETHERTYPE_IPV6);
168 ah = (struct arphdr *)req->hdata;
169 ah->ar_hrd = htons(ARPHRD_ETHER);
171 switch(ntohs(ah->ar_op)) {
172 case ARPOP_REVREQUEST:
174 etype = htons(ETHERTYPE_REVARP);
179 etype = htons(ETHERTYPE_ARP);
183 if (req->flags & IFENCAP_FLAG_BROADCAST)
184 lladdr = ifp->if_broadcastaddr;
187 return (EAFNOSUPPORT);
190 memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
191 memcpy(eh->ether_dhost, lladdr, ETHER_ADDR_LEN);
192 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
193 req->bufsize = sizeof(struct ether_header);
200 ether_resolve_addr(struct ifnet *ifp, struct mbuf *m,
201 const struct sockaddr *dst, struct route *ro, u_char *phdr,
202 uint32_t *pflags, struct llentry **plle)
204 struct ether_header *eh;
205 uint32_t lleflags = 0;
207 #if defined(INET) || defined(INET6)
213 eh = (struct ether_header *)phdr;
215 switch (dst->sa_family) {
218 if ((m->m_flags & (M_BCAST | M_MCAST)) == 0)
219 error = arpresolve(ifp, 0, m, dst, phdr, &lleflags,
222 if (m->m_flags & M_BCAST)
223 memcpy(eh->ether_dhost, ifp->if_broadcastaddr,
226 const struct in_addr *a;
227 a = &(((const struct sockaddr_in *)dst)->sin_addr);
228 ETHER_MAP_IP_MULTICAST(a, eh->ether_dhost);
230 etype = htons(ETHERTYPE_IP);
231 memcpy(&eh->ether_type, &etype, sizeof(etype));
232 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
238 if ((m->m_flags & M_MCAST) == 0)
239 error = nd6_resolve(ifp, 0, m, dst, phdr, &lleflags,
242 const struct in6_addr *a6;
243 a6 = &(((const struct sockaddr_in6 *)dst)->sin6_addr);
244 ETHER_MAP_IPV6_MULTICAST(a6, eh->ether_dhost);
245 etype = htons(ETHERTYPE_IPV6);
246 memcpy(&eh->ether_type, &etype, sizeof(etype));
247 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
252 if_printf(ifp, "can't handle af%d\n", dst->sa_family);
255 return (EAFNOSUPPORT);
258 if (error == EHOSTDOWN) {
259 if (ro != NULL && (ro->ro_flags & RT_HAS_GW) != 0)
260 error = EHOSTUNREACH;
266 *pflags = RT_MAY_LOOP;
267 if (lleflags & LLE_IFADDR)
274 * Ethernet output routine.
275 * Encapsulate a packet of type family for the local net.
276 * Use trailer local net encapsulation if enough data in first
277 * packet leaves a multiple of 512 bytes of data in remainder.
280 ether_output(struct ifnet *ifp, struct mbuf *m,
281 const struct sockaddr *dst, struct route *ro)
284 char linkhdr[ETHER_HDR_LEN], *phdr;
285 struct ether_header *eh;
288 int hlen; /* link layer header length */
290 struct llentry *lle = NULL;
296 /* XXX BPF uses ro_prepend */
297 if (ro->ro_prepend != NULL) {
298 phdr = ro->ro_prepend;
300 } else if (!(m->m_flags & (M_BCAST | M_MCAST))) {
301 if ((ro->ro_flags & RT_LLE_CACHE) != 0) {
304 (lle->la_flags & LLE_VALID) == 0) {
306 lle = NULL; /* redundant */
310 /* if we lookup, keep cache */
314 * Notify LLE code that
318 llentry_mark_used(lle);
321 phdr = lle->r_linkdata;
322 hlen = lle->r_hdrlen;
323 pflags = lle->r_flags;
329 error = mac_ifnet_check_transmit(ifp, m);
335 if (ifp->if_flags & IFF_MONITOR)
337 if (!((ifp->if_flags & IFF_UP) &&
338 (ifp->if_drv_flags & IFF_DRV_RUNNING)))
342 /* No prepend data supplied. Try to calculate ourselves. */
344 hlen = ETHER_HDR_LEN;
345 error = ether_resolve_addr(ifp, m, dst, ro, phdr, &pflags,
346 addref ? &lle : NULL);
347 if (addref && lle != NULL)
350 return (error == EWOULDBLOCK ? 0 : error);
353 if ((pflags & RT_L2_ME) != 0) {
354 update_mbuf_csumflags(m, m);
355 return (if_simloop(ifp, m, dst->sa_family, 0));
357 loop_copy = pflags & RT_MAY_LOOP;
360 * Add local net header. If no space in first mbuf,
363 * Note that we do prepend regardless of RT_HAS_HEADER flag.
364 * This is done because BPF code shifts m_data pointer
365 * to the end of ethernet header prior to calling if_output().
367 M_PREPEND(m, hlen, M_NOWAIT);
370 if ((pflags & RT_HAS_HEADER) == 0) {
371 eh = mtod(m, struct ether_header *);
372 memcpy(eh, phdr, hlen);
376 * If a simplex interface, and the packet is being sent to our
377 * Ethernet address or a broadcast address, loopback a copy.
378 * XXX To make a simplex device behave exactly like a duplex
379 * device, we should copy in the case of sending to our own
380 * ethernet address (thus letting the original actually appear
381 * on the wire). However, we don't do that here for security
382 * reasons and compatibility with the original behavior.
384 if ((m->m_flags & M_BCAST) && loop_copy && (ifp->if_flags & IFF_SIMPLEX) &&
385 ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
389 * Because if_simloop() modifies the packet, we need a
390 * writable copy through m_dup() instead of a readonly
391 * one as m_copy[m] would give us. The alternative would
392 * be to modify if_simloop() to handle the readonly mbuf,
393 * but performancewise it is mostly equivalent (trading
394 * extra data copying vs. extra locking).
396 * XXX This is a local workaround. A number of less
397 * often used kernel parts suffer from the same bug.
398 * See PR kern/105943 for a proposed general solution.
400 if ((n = m_dup(m, M_NOWAIT)) != NULL) {
401 update_mbuf_csumflags(m, n);
402 (void)if_simloop(ifp, n, dst->sa_family, hlen);
404 if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
408 * Bridges require special output handling.
410 if (ifp->if_bridge) {
411 BRIDGE_OUTPUT(ifp, m, error);
415 #if defined(INET) || defined(INET6)
417 (error = (*carp_output_p)(ifp, m, dst)))
421 /* Handle ng_ether(4) processing, if any */
422 if (ifp->if_l2com != NULL) {
423 KASSERT(ng_ether_output_p != NULL,
424 ("ng_ether_output_p is NULL"));
425 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
434 /* Continue with link-layer output */
435 return ether_output_frame(ifp, m);
439 ether_set_pcp(struct mbuf **mp, struct ifnet *ifp, uint8_t pcp)
441 struct ether_header *eh;
443 eh = mtod(*mp, struct ether_header *);
444 if (ntohs(eh->ether_type) == ETHERTYPE_VLAN ||
445 ether_8021q_frame(mp, ifp, ifp, 0, pcp))
447 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
452 * Ethernet link layer output routine to send a raw frame to the device.
454 * This assumes that the 14 byte Ethernet header is present and contiguous
455 * in the first mbuf (if BRIDGE'ing).
458 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
464 if (pcp != IFNET_PCP_NONE && ifp->if_type != IFT_L2VLAN &&
465 !ether_set_pcp(&m, ifp, pcp))
468 if (PFIL_HOOKED(&V_link_pfil_hook)) {
469 error = pfil_run_hooks(&V_link_pfil_hook, &m, ifp,
479 * Queue message on interface, update output statistics if
480 * successful, and start output if interface not yet active.
482 return ((ifp->if_transmit)(ifp, m));
486 * Process a received Ethernet packet; the packet is in the
487 * mbuf chain m with the ethernet header at the front.
490 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
492 struct ether_header *eh;
495 if ((ifp->if_flags & IFF_UP) == 0) {
500 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
501 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
506 if (m->m_len < ETHER_HDR_LEN) {
507 /* XXX maybe should pullup? */
508 if_printf(ifp, "discard frame w/o leading ethernet "
509 "header (len %u pkt len %u)\n",
510 m->m_len, m->m_pkthdr.len);
511 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
515 eh = mtod(m, struct ether_header *);
516 etype = ntohs(eh->ether_type);
517 random_harvest_queue_ether(m, sizeof(*m));
519 CURVNET_SET_QUIET(ifp->if_vnet);
521 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
522 if (ETHER_IS_BROADCAST(eh->ether_dhost))
523 m->m_flags |= M_BCAST;
525 m->m_flags |= M_MCAST;
526 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
531 * Tag the mbuf with an appropriate MAC label before any other
532 * consumers can get to it.
534 mac_ifnet_create_mbuf(ifp, m);
538 * Give bpf a chance at the packet.
540 ETHER_BPF_MTAP(ifp, m);
543 * If the CRC is still on the packet, trim it off. We do this once
544 * and once only in case we are re-entered. Nothing else on the
545 * Ethernet receive path expects to see the FCS.
547 if (m->m_flags & M_HASFCS) {
548 m_adj(m, -ETHER_CRC_LEN);
549 m->m_flags &= ~M_HASFCS;
552 if (!(ifp->if_capenable & IFCAP_HWSTATS))
553 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
555 /* Allow monitor mode to claim this frame, after stats are updated. */
556 if (ifp->if_flags & IFF_MONITOR) {
562 /* Handle input from a lagg(4) port */
563 if (ifp->if_type == IFT_IEEE8023ADLAG) {
564 KASSERT(lagg_input_p != NULL,
565 ("%s: if_lagg not loaded!", __func__));
566 m = (*lagg_input_p)(ifp, m);
568 ifp = m->m_pkthdr.rcvif;
576 * If the hardware did not process an 802.1Q tag, do this now,
577 * to allow 802.1P priority frames to be passed to the main input
579 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels.
581 if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) {
582 struct ether_vlan_header *evl;
584 if (m->m_len < sizeof(*evl) &&
585 (m = m_pullup(m, sizeof(*evl))) == NULL) {
587 if_printf(ifp, "cannot pullup VLAN header\n");
589 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
594 evl = mtod(m, struct ether_vlan_header *);
595 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
596 m->m_flags |= M_VLANTAG;
598 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
599 ETHER_HDR_LEN - ETHER_TYPE_LEN);
600 m_adj(m, ETHER_VLAN_ENCAP_LEN);
601 eh = mtod(m, struct ether_header *);
604 M_SETFIB(m, ifp->if_fib);
606 /* Allow ng_ether(4) to claim this frame. */
607 if (ifp->if_l2com != NULL) {
608 KASSERT(ng_ether_input_p != NULL,
609 ("%s: ng_ether_input_p is NULL", __func__));
610 m->m_flags &= ~M_PROMISC;
611 (*ng_ether_input_p)(ifp, &m);
616 eh = mtod(m, struct ether_header *);
620 * Allow if_bridge(4) to claim this frame.
621 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
622 * and the frame should be delivered locally.
624 if (ifp->if_bridge != NULL) {
625 m->m_flags &= ~M_PROMISC;
626 BRIDGE_INPUT(ifp, m);
631 eh = mtod(m, struct ether_header *);
634 #if defined(INET) || defined(INET6)
636 * Clear M_PROMISC on frame so that carp(4) will see it when the
637 * mbuf flows up to Layer 3.
638 * FreeBSD's implementation of carp(4) uses the inprotosw
639 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
640 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
641 * is outside the scope of the M_PROMISC test below.
642 * TODO: Maintain a hash table of ethernet addresses other than
643 * ether_dhost which may be active on this ifp.
645 if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
646 m->m_flags &= ~M_PROMISC;
651 * If the frame received was not for our MAC address, set the
652 * M_PROMISC flag on the mbuf chain. The frame may need to
653 * be seen by the rest of the Ethernet input path in case of
654 * re-entry (e.g. bridge, vlan, netgraph) but should not be
655 * seen by upper protocol layers.
657 if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
658 bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
659 m->m_flags |= M_PROMISC;
667 * Ethernet input dispatch; by default, direct dispatch here regardless of
668 * global configuration. However, if RSS is enabled, hook up RSS affinity
669 * so that when deferred or hybrid dispatch is enabled, we can redistribute
672 * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or
673 * not it had already done work distribution via multi-queue. Then we could
674 * direct dispatch in the event load balancing was already complete and
675 * handle the case of interfaces with different capabilities better.
677 * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions
678 * at multiple layers?
680 * XXXRW: For now, enable all this only if RSS is compiled in, although it
681 * works fine without RSS. Need to characterise the performance overhead
682 * of the detour through the netisr code in the event the result is always
686 ether_nh_input(struct mbuf *m)
690 KASSERT(m->m_pkthdr.rcvif != NULL,
691 ("%s: NULL interface pointer", __func__));
692 ether_input_internal(m->m_pkthdr.rcvif, m);
695 static struct netisr_handler ether_nh = {
697 .nh_handler = ether_nh_input,
698 .nh_proto = NETISR_ETHER,
700 .nh_policy = NETISR_POLICY_CPU,
701 .nh_dispatch = NETISR_DISPATCH_DIRECT,
702 .nh_m2cpuid = rss_m2cpuid,
704 .nh_policy = NETISR_POLICY_SOURCE,
705 .nh_dispatch = NETISR_DISPATCH_DIRECT,
710 ether_init(__unused void *arg)
713 netisr_register(ðer_nh);
715 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
718 vnet_ether_init(__unused void *arg)
722 /* Initialize packet filter hooks. */
723 V_link_pfil_hook.ph_type = PFIL_TYPE_AF;
724 V_link_pfil_hook.ph_af = AF_LINK;
725 if ((i = pfil_head_register(&V_link_pfil_hook)) != 0)
726 printf("%s: WARNING: unable to register pfil link hook, "
727 "error %d\n", __func__, i);
729 netisr_register_vnet(ðer_nh);
732 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY,
733 vnet_ether_init, NULL);
737 vnet_ether_pfil_destroy(__unused void *arg)
741 if ((i = pfil_head_unregister(&V_link_pfil_hook)) != 0)
742 printf("%s: WARNING: unable to unregister pfil link hook, "
743 "error %d\n", __func__, i);
745 VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY,
746 vnet_ether_pfil_destroy, NULL);
749 vnet_ether_destroy(__unused void *arg)
752 netisr_unregister_vnet(ðer_nh);
754 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY,
755 vnet_ether_destroy, NULL);
761 ether_input(struct ifnet *ifp, struct mbuf *m)
767 * The drivers are allowed to pass in a chain of packets linked with
768 * m_nextpkt. We split them up into separate packets here and pass
769 * them up. This allows the drivers to amortize the receive lock.
776 * We will rely on rcvif being set properly in the deferred context,
777 * so assert it is correct here.
779 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p "
780 "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp));
781 CURVNET_SET_QUIET(ifp->if_vnet);
782 netisr_dispatch(NETISR_ETHER, m);
789 * Upper layer processing for a received Ethernet packet.
792 ether_demux(struct ifnet *ifp, struct mbuf *m)
794 struct ether_header *eh;
798 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
800 /* Do not grab PROMISC frames in case we are re-entered. */
801 if (PFIL_HOOKED(&V_link_pfil_hook) && !(m->m_flags & M_PROMISC)) {
802 i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_IN, 0,
805 if (i != 0 || m == NULL)
809 eh = mtod(m, struct ether_header *);
810 ether_type = ntohs(eh->ether_type);
813 * If this frame has a VLAN tag other than 0, call vlan_input()
814 * if its module is loaded. Otherwise, drop.
816 if ((m->m_flags & M_VLANTAG) &&
817 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
818 if (ifp->if_vlantrunk == NULL) {
819 if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
823 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
825 /* Clear before possibly re-entering ether_input(). */
826 m->m_flags &= ~M_PROMISC;
827 (*vlan_input_p)(ifp, m);
832 * Pass promiscuously received frames to the upper layer if the user
833 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
835 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
841 * Reset layer specific mbuf flags to avoid confusing upper layers.
842 * Strip off Ethernet header.
844 m->m_flags &= ~M_VLANTAG;
846 m_adj(m, ETHER_HDR_LEN);
849 * Dispatch frame to upper layer.
851 switch (ether_type) {
858 if (ifp->if_flags & IFF_NOARP) {
859 /* Discard packet if ARP is disabled on interface */
874 netisr_dispatch(isr, m);
879 * Packet is to be discarded. If netgraph is present,
880 * hand the packet to it for last chance processing;
881 * otherwise dispose of it.
883 if (ifp->if_l2com != NULL) {
884 KASSERT(ng_ether_input_orphan_p != NULL,
885 ("ng_ether_input_orphan_p is NULL"));
887 * Put back the ethernet header so netgraph has a
888 * consistent view of inbound packets.
890 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
891 (*ng_ether_input_orphan_p)(ifp, m);
898 * Convert Ethernet address to printable (loggable) representation.
899 * This routine is for compatibility; it's better to just use
901 * printf("%6D", <pointer to address>, ":");
903 * since there's no static buffer involved.
906 ether_sprintf(const u_char *ap)
908 static char etherbuf[18];
909 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
914 * Perform common duties while attaching to interface list
917 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
921 struct sockaddr_dl *sdl;
923 ifp->if_addrlen = ETHER_ADDR_LEN;
924 ifp->if_hdrlen = ETHER_HDR_LEN;
926 ifp->if_mtu = ETHERMTU;
927 ifp->if_output = ether_output;
928 ifp->if_input = ether_input;
929 ifp->if_resolvemulti = ether_resolvemulti;
930 ifp->if_requestencap = ether_requestencap;
932 ifp->if_reassign = ether_reassign;
934 if (ifp->if_baudrate == 0)
935 ifp->if_baudrate = IF_Mbps(10); /* just a default */
936 ifp->if_broadcastaddr = etherbroadcastaddr;
939 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
940 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
941 sdl->sdl_type = IFT_ETHER;
942 sdl->sdl_alen = ifp->if_addrlen;
943 bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
945 if (ifp->if_hw_addr != NULL)
946 bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen);
948 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
949 if (ng_ether_attach_p != NULL)
950 (*ng_ether_attach_p)(ifp);
952 /* Announce Ethernet MAC address if non-zero. */
953 for (i = 0; i < ifp->if_addrlen; i++)
956 if (i != ifp->if_addrlen)
957 if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
959 uuid_ether_add(LLADDR(sdl));
961 /* Add necessary bits are setup; announce it now. */
962 EVENTHANDLER_INVOKE(ether_ifattach_event, ifp);
963 if (IS_DEFAULT_VNET(curvnet))
964 devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL);
968 * Perform common duties while detaching an Ethernet interface
971 ether_ifdetach(struct ifnet *ifp)
973 struct sockaddr_dl *sdl;
975 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr);
976 uuid_ether_del(LLADDR(sdl));
978 if (ifp->if_l2com != NULL) {
979 KASSERT(ng_ether_detach_p != NULL,
980 ("ng_ether_detach_p is NULL"));
981 (*ng_ether_detach_p)(ifp);
990 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused)
993 if (ifp->if_l2com != NULL) {
994 KASSERT(ng_ether_detach_p != NULL,
995 ("ng_ether_detach_p is NULL"));
996 (*ng_ether_detach_p)(ifp);
999 if (ng_ether_attach_p != NULL) {
1000 CURVNET_SET_QUIET(new_vnet);
1001 (*ng_ether_attach_p)(ifp);
1007 SYSCTL_DECL(_net_link);
1008 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
1012 * This is for reference. We have a table-driven version
1013 * of the little-endian crc32 generator, which is faster
1014 * than the double-loop.
1017 ether_crc32_le(const uint8_t *buf, size_t len)
1024 crc = 0xffffffff; /* initial value */
1026 for (i = 0; i < len; i++) {
1027 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1028 carry = (crc ^ data) & 1;
1031 crc = (crc ^ ETHER_CRC_POLY_LE);
1039 ether_crc32_le(const uint8_t *buf, size_t len)
1041 static const uint32_t crctab[] = {
1042 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1043 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1044 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1045 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1050 crc = 0xffffffff; /* initial value */
1052 for (i = 0; i < len; i++) {
1054 crc = (crc >> 4) ^ crctab[crc & 0xf];
1055 crc = (crc >> 4) ^ crctab[crc & 0xf];
1063 ether_crc32_be(const uint8_t *buf, size_t len)
1066 uint32_t crc, carry;
1070 crc = 0xffffffff; /* initial value */
1072 for (i = 0; i < len; i++) {
1073 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1074 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1077 crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1085 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1087 struct ifaddr *ifa = (struct ifaddr *) data;
1088 struct ifreq *ifr = (struct ifreq *) data;
1093 ifp->if_flags |= IFF_UP;
1095 switch (ifa->ifa_addr->sa_family) {
1098 ifp->if_init(ifp->if_softc); /* before arpwhohas */
1099 arp_ifinit(ifp, ifa);
1103 ifp->if_init(ifp->if_softc);
1109 bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0],
1115 * Set the interface MTU.
1117 if (ifr->ifr_mtu > ETHERMTU) {
1120 ifp->if_mtu = ifr->ifr_mtu;
1125 error = priv_check(curthread, PRIV_NET_SETLANPCP);
1128 if (ifr->ifr_lan_pcp > 7 &&
1129 ifr->ifr_lan_pcp != IFNET_PCP_NONE) {
1132 ifp->if_pcp = ifr->ifr_lan_pcp;
1133 /* broadcast event about PCP change */
1134 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP);
1139 ifr->ifr_lan_pcp = ifp->if_pcp;
1143 error = EINVAL; /* XXX netbsd has ENOTTY??? */
1150 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1151 struct sockaddr *sa)
1153 struct sockaddr_dl *sdl;
1155 struct sockaddr_in *sin;
1158 struct sockaddr_in6 *sin6;
1162 switch(sa->sa_family) {
1165 * No mapping needed. Just check that it's a valid MC address.
1167 sdl = (struct sockaddr_dl *)sa;
1168 e_addr = LLADDR(sdl);
1169 if (!ETHER_IS_MULTICAST(e_addr))
1170 return EADDRNOTAVAIL;
1176 sin = (struct sockaddr_in *)sa;
1177 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1178 return EADDRNOTAVAIL;
1179 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1180 sdl->sdl_alen = ETHER_ADDR_LEN;
1181 e_addr = LLADDR(sdl);
1182 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1183 *llsa = (struct sockaddr *)sdl;
1188 sin6 = (struct sockaddr_in6 *)sa;
1189 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1191 * An IP6 address of 0 means listen to all
1192 * of the Ethernet multicast address used for IP6.
1193 * (This is used for multicast routers.)
1195 ifp->if_flags |= IFF_ALLMULTI;
1199 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1200 return EADDRNOTAVAIL;
1201 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1202 sdl->sdl_alen = ETHER_ADDR_LEN;
1203 e_addr = LLADDR(sdl);
1204 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1205 *llsa = (struct sockaddr *)sdl;
1211 * Well, the text isn't quite right, but it's the name
1214 return EAFNOSUPPORT;
1218 static moduledata_t ether_mod = {
1223 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1225 struct ether_vlan_header vlan;
1228 KASSERT((m->m_flags & M_VLANTAG) != 0,
1229 ("%s: vlan information not present", __func__));
1230 KASSERT(m->m_len >= sizeof(struct ether_header),
1231 ("%s: mbuf not large enough for header", __func__));
1232 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1233 vlan.evl_proto = vlan.evl_encap_proto;
1234 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1235 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1236 m->m_len -= sizeof(struct ether_header);
1237 m->m_data += sizeof(struct ether_header);
1239 * If a data link has been supplied by the caller, then we will need to
1240 * re-create a stack allocated mbuf chain with the following structure:
1242 * (1) mbuf #1 will contain the supplied data link
1243 * (2) mbuf #2 will contain the vlan header
1244 * (3) mbuf #3 will contain the original mbuf's packet data
1246 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1250 mv.m_data = (caddr_t)&vlan;
1251 mv.m_len = sizeof(vlan);
1257 bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1258 m->m_len += sizeof(struct ether_header);
1259 m->m_data -= sizeof(struct ether_header);
1263 ether_vlanencap(struct mbuf *m, uint16_t tag)
1265 struct ether_vlan_header *evl;
1267 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT);
1270 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1272 if (m->m_len < sizeof(*evl)) {
1273 m = m_pullup(m, sizeof(*evl));
1279 * Transform the Ethernet header into an Ethernet header
1280 * with 802.1Q encapsulation.
1282 evl = mtod(m, struct ether_vlan_header *);
1283 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1284 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1285 evl->evl_encap_proto = htons(ETHERTYPE_VLAN);
1286 evl->evl_tag = htons(tag);
1290 static SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW, 0,
1291 "IEEE 802.1Q VLAN");
1292 static SYSCTL_NODE(_net_link_vlan, PF_LINK, link, CTLFLAG_RW, 0,
1295 VNET_DEFINE_STATIC(int, soft_pad);
1296 #define V_soft_pad VNET(soft_pad)
1297 SYSCTL_INT(_net_link_vlan, OID_AUTO, soft_pad, CTLFLAG_RW | CTLFLAG_VNET,
1298 &VNET_NAME(soft_pad), 0,
1299 "pad short frames before tagging");
1302 * For now, make preserving PCP via an mbuf tag optional, as it increases
1303 * per-packet memory allocations and frees. In the future, it would be
1304 * preferable to reuse ether_vtag for this, or similar.
1306 int vlan_mtag_pcp = 0;
1307 SYSCTL_INT(_net_link_vlan, OID_AUTO, mtag_pcp, CTLFLAG_RW,
1309 "Retain VLAN PCP information as packets are passed up the stack");
1312 ether_8021q_frame(struct mbuf **mp, struct ifnet *ife, struct ifnet *p,
1313 uint16_t vid, uint8_t pcp)
1318 static const char pad[8]; /* just zeros */
1321 * Pad the frame to the minimum size allowed if told to.
1322 * This option is in accord with IEEE Std 802.1Q, 2003 Ed.,
1323 * paragraph C.4.4.3.b. It can help to work around buggy
1324 * bridges that violate paragraph C.4.4.3.a from the same
1325 * document, i.e., fail to pad short frames after untagging.
1326 * E.g., a tagged frame 66 bytes long (incl. FCS) is OK, but
1327 * untagging it will produce a 62-byte frame, which is a runt
1328 * and requires padding. There are VLAN-enabled network
1329 * devices that just discard such runts instead or mishandle
1332 if (V_soft_pad && p->if_type == IFT_ETHER) {
1333 for (n = ETHERMIN + ETHER_HDR_LEN - (*mp)->m_pkthdr.len;
1334 n > 0; n -= sizeof(pad)) {
1335 if (!m_append(*mp, min(n, sizeof(pad)), pad))
1341 if_printf(ife, "cannot pad short frame");
1347 * If underlying interface can do VLAN tag insertion itself,
1348 * just pass the packet along. However, we need some way to
1349 * tell the interface where the packet came from so that it
1350 * knows how to find the VLAN tag to use, so we attach a
1351 * packet tag that holds it.
1353 if (vlan_mtag_pcp && (mtag = m_tag_locate(*mp, MTAG_8021Q,
1354 MTAG_8021Q_PCP_OUT, NULL)) != NULL)
1355 tag = EVL_MAKETAG(vid, *(uint8_t *)(mtag + 1), 0);
1357 tag = EVL_MAKETAG(vid, pcp, 0);
1358 if (p->if_capenable & IFCAP_VLAN_HWTAGGING) {
1359 (*mp)->m_pkthdr.ether_vtag = tag;
1360 (*mp)->m_flags |= M_VLANTAG;
1362 *mp = ether_vlanencap(*mp, tag);
1364 if_printf(ife, "unable to prepend 802.1Q header");
1371 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1372 MODULE_VERSION(ether, 1);