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>
43 #include <sys/devctl.h>
44 #include <sys/eventhandler.h>
46 #include <sys/kernel.h>
48 #include <sys/malloc.h>
49 #include <sys/module.h>
53 #include <sys/random.h>
54 #include <sys/socket.h>
55 #include <sys/sockio.h>
56 #include <sys/sysctl.h>
59 #include <net/ieee_oui.h>
61 #include <net/if_var.h>
62 #include <net/if_arp.h>
63 #include <net/netisr.h>
64 #include <net/route.h>
65 #include <net/if_llc.h>
66 #include <net/if_dl.h>
67 #include <net/if_types.h>
69 #include <net/ethernet.h>
70 #include <net/if_bridgevar.h>
71 #include <net/if_vlan_var.h>
72 #include <net/if_llatbl.h>
74 #include <net/rss_config.h>
77 #include <netpfil/pf/pf_mtag.h>
79 #if defined(INET) || defined(INET6)
80 #include <netinet/in.h>
81 #include <netinet/in_var.h>
82 #include <netinet/if_ether.h>
83 #include <netinet/ip_carp.h>
84 #include <netinet/ip_var.h>
87 #include <netinet6/nd6.h>
89 #include <security/mac/mac_framework.h>
91 #include <crypto/sha1.h>
94 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2);
95 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN);
98 VNET_DEFINE(pfil_head_t, link_pfil_head); /* Packet filter hooks */
100 /* netgraph node hooks for ng_ether(4) */
101 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
102 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
103 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
104 void (*ng_ether_attach_p)(struct ifnet *ifp);
105 void (*ng_ether_detach_p)(struct ifnet *ifp);
107 void (*vlan_input_p)(struct ifnet *, struct mbuf *);
109 /* if_bridge(4) support */
110 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
112 /* if_lagg(4) support */
113 struct mbuf *(*lagg_input_ethernet_p)(struct ifnet *, struct mbuf *);
115 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
116 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
118 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
120 static int ether_requestencap(struct ifnet *, struct if_encap_req *);
122 #define senderr(e) do { error = (e); goto bad;} while (0)
125 update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst)
129 if (src->m_pkthdr.csum_flags & CSUM_IP)
130 csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
131 if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
132 csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
133 if (src->m_pkthdr.csum_flags & CSUM_SCTP)
134 csum_flags |= CSUM_SCTP_VALID;
135 dst->m_pkthdr.csum_flags |= csum_flags;
136 if (csum_flags & CSUM_DATA_VALID)
137 dst->m_pkthdr.csum_data = 0xffff;
141 * Handle link-layer encapsulation requests.
144 ether_requestencap(struct ifnet *ifp, struct if_encap_req *req)
146 struct ether_header *eh;
149 const u_char *lladdr;
151 if (req->rtype != IFENCAP_LL)
154 if (req->bufsize < ETHER_HDR_LEN)
157 eh = (struct ether_header *)req->buf;
158 lladdr = req->lladdr;
161 switch (req->family) {
163 etype = htons(ETHERTYPE_IP);
166 etype = htons(ETHERTYPE_IPV6);
169 ah = (struct arphdr *)req->hdata;
170 ah->ar_hrd = htons(ARPHRD_ETHER);
172 switch(ntohs(ah->ar_op)) {
173 case ARPOP_REVREQUEST:
175 etype = htons(ETHERTYPE_REVARP);
180 etype = htons(ETHERTYPE_ARP);
184 if (req->flags & IFENCAP_FLAG_BROADCAST)
185 lladdr = ifp->if_broadcastaddr;
188 return (EAFNOSUPPORT);
191 memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
192 memcpy(eh->ether_dhost, lladdr, ETHER_ADDR_LEN);
193 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
194 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 int af = RO_GET_FAMILY(ro, dst);
240 error = nd6_resolve(ifp, LLE_SF(af, 0), m, dst, phdr,
243 const struct in6_addr *a6;
244 a6 = &(((const struct sockaddr_in6 *)dst)->sin6_addr);
245 ETHER_MAP_IPV6_MULTICAST(a6, eh->ether_dhost);
246 etype = htons(ETHERTYPE_IPV6);
247 memcpy(&eh->ether_type, &etype, sizeof(etype));
248 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
253 if_printf(ifp, "can't handle af%d\n", dst->sa_family);
256 return (EAFNOSUPPORT);
259 if (error == EHOSTDOWN) {
260 if (ro != NULL && (ro->ro_flags & RT_HAS_GW) != 0)
261 error = EHOSTUNREACH;
267 *pflags = RT_MAY_LOOP;
268 if (lleflags & LLE_IFADDR)
275 * Ethernet output routine.
276 * Encapsulate a packet of type family for the local net.
277 * Use trailer local net encapsulation if enough data in first
278 * packet leaves a multiple of 512 bytes of data in remainder.
281 ether_output(struct ifnet *ifp, struct mbuf *m,
282 const struct sockaddr *dst, struct route *ro)
285 char linkhdr[ETHER_HDR_LEN], *phdr;
286 struct ether_header *eh;
289 int hlen; /* link layer header length */
291 struct llentry *lle = NULL;
297 /* XXX BPF uses ro_prepend */
298 if (ro->ro_prepend != NULL) {
299 phdr = ro->ro_prepend;
301 } else if (!(m->m_flags & (M_BCAST | M_MCAST))) {
302 if ((ro->ro_flags & RT_LLE_CACHE) != 0) {
305 (lle->la_flags & LLE_VALID) == 0) {
307 lle = NULL; /* redundant */
311 /* if we lookup, keep cache */
315 * Notify LLE code that
319 llentry_provide_feedback(lle);
322 phdr = lle->r_linkdata;
323 hlen = lle->r_hdrlen;
324 pflags = lle->r_flags;
330 error = mac_ifnet_check_transmit(ifp, m);
336 if (ifp->if_flags & IFF_MONITOR)
338 if (!((ifp->if_flags & IFF_UP) &&
339 (ifp->if_drv_flags & IFF_DRV_RUNNING)))
343 /* No prepend data supplied. Try to calculate ourselves. */
345 hlen = ETHER_HDR_LEN;
346 error = ether_resolve_addr(ifp, m, dst, ro, phdr, &pflags,
347 addref ? &lle : NULL);
348 if (addref && lle != NULL)
351 return (error == EWOULDBLOCK ? 0 : error);
354 if ((pflags & RT_L2_ME) != 0) {
355 update_mbuf_csumflags(m, m);
356 return (if_simloop(ifp, m, RO_GET_FAMILY(ro, dst), 0));
358 loop_copy = (pflags & RT_MAY_LOOP) != 0;
361 * Add local net header. If no space in first mbuf,
364 * Note that we do prepend regardless of RT_HAS_HEADER flag.
365 * This is done because BPF code shifts m_data pointer
366 * to the end of ethernet header prior to calling if_output().
368 M_PREPEND(m, hlen, M_NOWAIT);
371 if ((pflags & RT_HAS_HEADER) == 0) {
372 eh = mtod(m, struct ether_header *);
373 memcpy(eh, phdr, hlen);
377 * If a simplex interface, and the packet is being sent to our
378 * Ethernet address or a broadcast address, loopback a copy.
379 * XXX To make a simplex device behave exactly like a duplex
380 * device, we should copy in the case of sending to our own
381 * ethernet address (thus letting the original actually appear
382 * on the wire). However, we don't do that here for security
383 * reasons and compatibility with the original behavior.
385 if ((m->m_flags & M_BCAST) && loop_copy && (ifp->if_flags & IFF_SIMPLEX) &&
386 ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
390 * Because if_simloop() modifies the packet, we need a
391 * writable copy through m_dup() instead of a readonly
392 * one as m_copy[m] would give us. The alternative would
393 * be to modify if_simloop() to handle the readonly mbuf,
394 * but performancewise it is mostly equivalent (trading
395 * extra data copying vs. extra locking).
397 * XXX This is a local workaround. A number of less
398 * often used kernel parts suffer from the same bug.
399 * See PR kern/105943 for a proposed general solution.
401 if ((n = m_dup(m, M_NOWAIT)) != NULL) {
402 update_mbuf_csumflags(m, n);
403 (void)if_simloop(ifp, n, RO_GET_FAMILY(ro, dst), hlen);
405 if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
409 * Bridges require special output handling.
411 if (ifp->if_bridge) {
412 BRIDGE_OUTPUT(ifp, m, error);
416 #if defined(INET) || defined(INET6)
418 (error = (*carp_output_p)(ifp, m, dst)))
422 /* Handle ng_ether(4) processing, if any */
423 if (ifp->if_l2com != NULL) {
424 KASSERT(ng_ether_output_p != NULL,
425 ("ng_ether_output_p is NULL"));
426 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
435 /* Continue with link-layer output */
436 return ether_output_frame(ifp, m);
440 ether_set_pcp(struct mbuf **mp, struct ifnet *ifp, uint8_t pcp)
442 struct ether_8021q_tag qtag;
443 struct ether_header *eh;
445 eh = mtod(*mp, struct ether_header *);
446 if (ntohs(eh->ether_type) == ETHERTYPE_VLAN ||
447 ntohs(eh->ether_type) == ETHERTYPE_QINQ)
452 qtag.proto = ETHERTYPE_VLAN;
453 if (ether_8021q_frame(mp, ifp, ifp, &qtag))
455 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
460 * Ethernet link layer output routine to send a raw frame to the device.
462 * This assumes that the 14 byte Ethernet header is present and contiguous
463 * in the first mbuf (if BRIDGE'ing).
466 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
471 if (pcp != IFNET_PCP_NONE && ifp->if_type != IFT_L2VLAN &&
472 !ether_set_pcp(&m, ifp, pcp))
475 if (PFIL_HOOKED_OUT(V_link_pfil_head))
476 switch (pfil_run_hooks(V_link_pfil_head, &m, ifp, PFIL_OUT,
485 #if defined(INET6) && defined(INET)
486 /* draft-ietf-6man-ipv6only-flag */
487 /* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */
488 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) {
489 struct ether_header *eh;
491 eh = mtod(m, struct ether_header *);
492 switch (ntohs(eh->ether_type)) {
495 case ETHERTYPE_REVARP:
497 return (EAFNOSUPPORT);
506 * Queue message on interface, update output statistics if
507 * successful, and start output if interface not yet active.
509 return ((ifp->if_transmit)(ifp, m));
513 * Process a received Ethernet packet; the packet is in the
514 * mbuf chain m with the ethernet header at the front.
517 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
519 struct ether_header *eh;
522 if ((ifp->if_flags & IFF_UP) == 0) {
527 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
528 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
533 if (m->m_len < ETHER_HDR_LEN) {
534 /* XXX maybe should pullup? */
535 if_printf(ifp, "discard frame w/o leading ethernet "
536 "header (len %u pkt len %u)\n",
537 m->m_len, m->m_pkthdr.len);
538 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
542 eh = mtod(m, struct ether_header *);
543 etype = ntohs(eh->ether_type);
544 random_harvest_queue_ether(m, sizeof(*m));
547 #if defined(INET6) && defined(INET)
548 /* draft-ietf-6man-ipv6only-flag */
549 /* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */
550 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) {
554 case ETHERTYPE_REVARP:
564 CURVNET_SET_QUIET(ifp->if_vnet);
566 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
567 if (ETHER_IS_BROADCAST(eh->ether_dhost))
568 m->m_flags |= M_BCAST;
570 m->m_flags |= M_MCAST;
571 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
576 * Tag the mbuf with an appropriate MAC label before any other
577 * consumers can get to it.
579 mac_ifnet_create_mbuf(ifp, m);
583 * Give bpf a chance at the packet.
585 ETHER_BPF_MTAP(ifp, m);
588 * If the CRC is still on the packet, trim it off. We do this once
589 * and once only in case we are re-entered. Nothing else on the
590 * Ethernet receive path expects to see the FCS.
592 if (m->m_flags & M_HASFCS) {
593 m_adj(m, -ETHER_CRC_LEN);
594 m->m_flags &= ~M_HASFCS;
597 if (!(ifp->if_capenable & IFCAP_HWSTATS))
598 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
600 /* Allow monitor mode to claim this frame, after stats are updated. */
601 if (ifp->if_flags & IFF_MONITOR) {
607 /* Handle input from a lagg(4) port */
608 if (ifp->if_type == IFT_IEEE8023ADLAG) {
609 KASSERT(lagg_input_ethernet_p != NULL,
610 ("%s: if_lagg not loaded!", __func__));
611 m = (*lagg_input_ethernet_p)(ifp, m);
613 ifp = m->m_pkthdr.rcvif;
621 * If the hardware did not process an 802.1Q tag, do this now,
622 * to allow 802.1P priority frames to be passed to the main input
625 if ((m->m_flags & M_VLANTAG) == 0 &&
626 ((etype == ETHERTYPE_VLAN) || (etype == ETHERTYPE_QINQ))) {
627 struct ether_vlan_header *evl;
629 if (m->m_len < sizeof(*evl) &&
630 (m = m_pullup(m, sizeof(*evl))) == NULL) {
632 if_printf(ifp, "cannot pullup VLAN header\n");
634 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
639 evl = mtod(m, struct ether_vlan_header *);
640 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
641 m->m_flags |= M_VLANTAG;
643 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
644 ETHER_HDR_LEN - ETHER_TYPE_LEN);
645 m_adj(m, ETHER_VLAN_ENCAP_LEN);
646 eh = mtod(m, struct ether_header *);
649 M_SETFIB(m, ifp->if_fib);
651 /* Allow ng_ether(4) to claim this frame. */
652 if (ifp->if_l2com != NULL) {
653 KASSERT(ng_ether_input_p != NULL,
654 ("%s: ng_ether_input_p is NULL", __func__));
655 m->m_flags &= ~M_PROMISC;
656 (*ng_ether_input_p)(ifp, &m);
661 eh = mtod(m, struct ether_header *);
665 * Allow if_bridge(4) to claim this frame.
666 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
667 * and the frame should be delivered locally.
669 if (ifp->if_bridge != NULL) {
670 m->m_flags &= ~M_PROMISC;
671 BRIDGE_INPUT(ifp, m);
676 eh = mtod(m, struct ether_header *);
679 #if defined(INET) || defined(INET6)
681 * Clear M_PROMISC on frame so that carp(4) will see it when the
682 * mbuf flows up to Layer 3.
683 * FreeBSD's implementation of carp(4) uses the inprotosw
684 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
685 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
686 * is outside the scope of the M_PROMISC test below.
687 * TODO: Maintain a hash table of ethernet addresses other than
688 * ether_dhost which may be active on this ifp.
690 if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
691 m->m_flags &= ~M_PROMISC;
696 * If the frame received was not for our MAC address, set the
697 * M_PROMISC flag on the mbuf chain. The frame may need to
698 * be seen by the rest of the Ethernet input path in case of
699 * re-entry (e.g. bridge, vlan, netgraph) but should not be
700 * seen by upper protocol layers.
702 if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
703 bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
704 m->m_flags |= M_PROMISC;
712 * Ethernet input dispatch; by default, direct dispatch here regardless of
713 * global configuration. However, if RSS is enabled, hook up RSS affinity
714 * so that when deferred or hybrid dispatch is enabled, we can redistribute
717 * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or
718 * not it had already done work distribution via multi-queue. Then we could
719 * direct dispatch in the event load balancing was already complete and
720 * handle the case of interfaces with different capabilities better.
722 * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions
723 * at multiple layers?
725 * XXXRW: For now, enable all this only if RSS is compiled in, although it
726 * works fine without RSS. Need to characterise the performance overhead
727 * of the detour through the netisr code in the event the result is always
731 ether_nh_input(struct mbuf *m)
735 KASSERT(m->m_pkthdr.rcvif != NULL,
736 ("%s: NULL interface pointer", __func__));
737 ether_input_internal(m->m_pkthdr.rcvif, m);
740 static struct netisr_handler ether_nh = {
742 .nh_handler = ether_nh_input,
743 .nh_proto = NETISR_ETHER,
745 .nh_policy = NETISR_POLICY_CPU,
746 .nh_dispatch = NETISR_DISPATCH_DIRECT,
747 .nh_m2cpuid = rss_m2cpuid,
749 .nh_policy = NETISR_POLICY_SOURCE,
750 .nh_dispatch = NETISR_DISPATCH_DIRECT,
755 ether_init(__unused void *arg)
758 netisr_register(ðer_nh);
760 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
763 vnet_ether_init(__unused void *arg)
765 struct pfil_head_args args;
767 args.pa_version = PFIL_VERSION;
768 args.pa_flags = PFIL_IN | PFIL_OUT;
769 args.pa_type = PFIL_TYPE_ETHERNET;
770 args.pa_headname = PFIL_ETHER_NAME;
771 V_link_pfil_head = pfil_head_register(&args);
774 netisr_register_vnet(ðer_nh);
777 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY,
778 vnet_ether_init, NULL);
782 vnet_ether_pfil_destroy(__unused void *arg)
785 pfil_head_unregister(V_link_pfil_head);
787 VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY,
788 vnet_ether_pfil_destroy, NULL);
791 vnet_ether_destroy(__unused void *arg)
794 netisr_unregister_vnet(ðer_nh);
796 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY,
797 vnet_ether_destroy, NULL);
801 ether_input(struct ifnet *ifp, struct mbuf *m)
803 struct epoch_tracker et;
807 needs_epoch = !(ifp->if_flags & IFF_KNOWSEPOCH);
810 * The drivers are allowed to pass in a chain of packets linked with
811 * m_nextpkt. We split them up into separate packets here and pass
812 * them up. This allows the drivers to amortize the receive lock.
814 CURVNET_SET_QUIET(ifp->if_vnet);
815 if (__predict_false(needs_epoch))
822 * We will rely on rcvif being set properly in the deferred
823 * context, so assert it is correct here.
825 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
826 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p "
827 "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp));
828 netisr_dispatch(NETISR_ETHER, m);
831 if (__predict_false(needs_epoch))
837 * Upper layer processing for a received Ethernet packet.
840 ether_demux(struct ifnet *ifp, struct mbuf *m)
842 struct ether_header *eh;
847 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
849 /* Do not grab PROMISC frames in case we are re-entered. */
850 if (PFIL_HOOKED_IN(V_link_pfil_head) && !(m->m_flags & M_PROMISC)) {
851 i = pfil_run_hooks(V_link_pfil_head, &m, ifp, PFIL_IN, NULL);
852 if (i != 0 || m == NULL)
856 eh = mtod(m, struct ether_header *);
857 ether_type = ntohs(eh->ether_type);
860 * If this frame has a VLAN tag other than 0, call vlan_input()
861 * if its module is loaded. Otherwise, drop.
863 if ((m->m_flags & M_VLANTAG) &&
864 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
865 if (ifp->if_vlantrunk == NULL) {
866 if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
870 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
872 /* Clear before possibly re-entering ether_input(). */
873 m->m_flags &= ~M_PROMISC;
874 (*vlan_input_p)(ifp, m);
879 * Pass promiscuously received frames to the upper layer if the user
880 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
882 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
888 * Reset layer specific mbuf flags to avoid confusing upper layers.
889 * Strip off Ethernet header.
891 m->m_flags &= ~M_VLANTAG;
893 m_adj(m, ETHER_HDR_LEN);
896 * Dispatch frame to upper layer.
898 switch (ether_type) {
905 if (ifp->if_flags & IFF_NOARP) {
906 /* Discard packet if ARP is disabled on interface */
921 netisr_dispatch(isr, m);
926 * Packet is to be discarded. If netgraph is present,
927 * hand the packet to it for last chance processing;
928 * otherwise dispose of it.
930 if (ifp->if_l2com != NULL) {
931 KASSERT(ng_ether_input_orphan_p != NULL,
932 ("ng_ether_input_orphan_p is NULL"));
934 * Put back the ethernet header so netgraph has a
935 * consistent view of inbound packets.
937 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
938 (*ng_ether_input_orphan_p)(ifp, m);
945 * Convert Ethernet address to printable (loggable) representation.
946 * This routine is for compatibility; it's better to just use
948 * printf("%6D", <pointer to address>, ":");
950 * since there's no static buffer involved.
953 ether_sprintf(const u_char *ap)
955 static char etherbuf[18];
956 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
961 * Perform common duties while attaching to interface list
964 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
968 struct sockaddr_dl *sdl;
970 ifp->if_addrlen = ETHER_ADDR_LEN;
971 ifp->if_hdrlen = ETHER_HDR_LEN;
972 ifp->if_mtu = ETHERMTU;
974 ifp->if_output = ether_output;
975 ifp->if_input = ether_input;
976 ifp->if_resolvemulti = ether_resolvemulti;
977 ifp->if_requestencap = ether_requestencap;
979 ifp->if_reassign = ether_reassign;
981 if (ifp->if_baudrate == 0)
982 ifp->if_baudrate = IF_Mbps(10); /* just a default */
983 ifp->if_broadcastaddr = etherbroadcastaddr;
986 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
987 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
988 sdl->sdl_type = IFT_ETHER;
989 sdl->sdl_alen = ifp->if_addrlen;
990 bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
992 if (ifp->if_hw_addr != NULL)
993 bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen);
995 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
996 if (ng_ether_attach_p != NULL)
997 (*ng_ether_attach_p)(ifp);
999 /* Announce Ethernet MAC address if non-zero. */
1000 for (i = 0; i < ifp->if_addrlen; i++)
1003 if (i != ifp->if_addrlen)
1004 if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
1006 uuid_ether_add(LLADDR(sdl));
1008 /* Add necessary bits are setup; announce it now. */
1009 EVENTHANDLER_INVOKE(ether_ifattach_event, ifp);
1010 if (IS_DEFAULT_VNET(curvnet))
1011 devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL);
1015 * Perform common duties while detaching an Ethernet interface
1018 ether_ifdetach(struct ifnet *ifp)
1020 struct sockaddr_dl *sdl;
1022 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr);
1023 uuid_ether_del(LLADDR(sdl));
1025 if (ifp->if_l2com != NULL) {
1026 KASSERT(ng_ether_detach_p != NULL,
1027 ("ng_ether_detach_p is NULL"));
1028 (*ng_ether_detach_p)(ifp);
1037 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused)
1040 if (ifp->if_l2com != NULL) {
1041 KASSERT(ng_ether_detach_p != NULL,
1042 ("ng_ether_detach_p is NULL"));
1043 (*ng_ether_detach_p)(ifp);
1046 if (ng_ether_attach_p != NULL) {
1047 CURVNET_SET_QUIET(new_vnet);
1048 (*ng_ether_attach_p)(ifp);
1054 SYSCTL_DECL(_net_link);
1055 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1060 * This is for reference. We have a table-driven version
1061 * of the little-endian crc32 generator, which is faster
1062 * than the double-loop.
1065 ether_crc32_le(const uint8_t *buf, size_t len)
1072 crc = 0xffffffff; /* initial value */
1074 for (i = 0; i < len; i++) {
1075 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1076 carry = (crc ^ data) & 1;
1079 crc = (crc ^ ETHER_CRC_POLY_LE);
1087 ether_crc32_le(const uint8_t *buf, size_t len)
1089 static const uint32_t crctab[] = {
1090 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1091 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1092 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1093 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1098 crc = 0xffffffff; /* initial value */
1100 for (i = 0; i < len; i++) {
1102 crc = (crc >> 4) ^ crctab[crc & 0xf];
1103 crc = (crc >> 4) ^ crctab[crc & 0xf];
1111 ether_crc32_be(const uint8_t *buf, size_t len)
1114 uint32_t crc, carry;
1118 crc = 0xffffffff; /* initial value */
1120 for (i = 0; i < len; i++) {
1121 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1122 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1125 crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1133 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1135 struct ifaddr *ifa = (struct ifaddr *) data;
1136 struct ifreq *ifr = (struct ifreq *) data;
1141 ifp->if_flags |= IFF_UP;
1143 switch (ifa->ifa_addr->sa_family) {
1146 ifp->if_init(ifp->if_softc); /* before arpwhohas */
1147 arp_ifinit(ifp, ifa);
1151 ifp->if_init(ifp->if_softc);
1157 bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0],
1163 * Set the interface MTU.
1165 if (ifr->ifr_mtu > ETHERMTU) {
1168 ifp->if_mtu = ifr->ifr_mtu;
1173 error = priv_check(curthread, PRIV_NET_SETLANPCP);
1176 if (ifr->ifr_lan_pcp > 7 &&
1177 ifr->ifr_lan_pcp != IFNET_PCP_NONE) {
1180 ifp->if_pcp = ifr->ifr_lan_pcp;
1181 /* broadcast event about PCP change */
1182 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP);
1187 ifr->ifr_lan_pcp = ifp->if_pcp;
1191 error = EINVAL; /* XXX netbsd has ENOTTY??? */
1198 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1199 struct sockaddr *sa)
1201 struct sockaddr_dl *sdl;
1203 struct sockaddr_in *sin;
1206 struct sockaddr_in6 *sin6;
1210 switch(sa->sa_family) {
1213 * No mapping needed. Just check that it's a valid MC address.
1215 sdl = (struct sockaddr_dl *)sa;
1216 e_addr = LLADDR(sdl);
1217 if (!ETHER_IS_MULTICAST(e_addr))
1218 return EADDRNOTAVAIL;
1224 sin = (struct sockaddr_in *)sa;
1225 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1226 return EADDRNOTAVAIL;
1227 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1228 sdl->sdl_alen = ETHER_ADDR_LEN;
1229 e_addr = LLADDR(sdl);
1230 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1231 *llsa = (struct sockaddr *)sdl;
1236 sin6 = (struct sockaddr_in6 *)sa;
1237 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1239 * An IP6 address of 0 means listen to all
1240 * of the Ethernet multicast address used for IP6.
1241 * (This is used for multicast routers.)
1243 ifp->if_flags |= IFF_ALLMULTI;
1247 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1248 return EADDRNOTAVAIL;
1249 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1250 sdl->sdl_alen = ETHER_ADDR_LEN;
1251 e_addr = LLADDR(sdl);
1252 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1253 *llsa = (struct sockaddr *)sdl;
1259 * Well, the text isn't quite right, but it's the name
1262 return EAFNOSUPPORT;
1266 static moduledata_t ether_mod = {
1271 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1273 struct ether_vlan_header vlan;
1276 KASSERT((m->m_flags & M_VLANTAG) != 0,
1277 ("%s: vlan information not present", __func__));
1278 KASSERT(m->m_len >= sizeof(struct ether_header),
1279 ("%s: mbuf not large enough for header", __func__));
1280 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1281 vlan.evl_proto = vlan.evl_encap_proto;
1282 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1283 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1284 m->m_len -= sizeof(struct ether_header);
1285 m->m_data += sizeof(struct ether_header);
1287 * If a data link has been supplied by the caller, then we will need to
1288 * re-create a stack allocated mbuf chain with the following structure:
1290 * (1) mbuf #1 will contain the supplied data link
1291 * (2) mbuf #2 will contain the vlan header
1292 * (3) mbuf #3 will contain the original mbuf's packet data
1294 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1298 mv.m_data = (caddr_t)&vlan;
1299 mv.m_len = sizeof(vlan);
1305 bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1306 m->m_len += sizeof(struct ether_header);
1307 m->m_data -= sizeof(struct ether_header);
1311 ether_vlanencap_proto(struct mbuf *m, uint16_t tag, uint16_t proto)
1313 struct ether_vlan_header *evl;
1315 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT);
1318 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1320 if (m->m_len < sizeof(*evl)) {
1321 m = m_pullup(m, sizeof(*evl));
1327 * Transform the Ethernet header into an Ethernet header
1328 * with 802.1Q encapsulation.
1330 evl = mtod(m, struct ether_vlan_header *);
1331 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1332 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1333 evl->evl_encap_proto = htons(proto);
1334 evl->evl_tag = htons(tag);
1338 static SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1339 "IEEE 802.1Q VLAN");
1340 static SYSCTL_NODE(_net_link_vlan, PF_LINK, link,
1341 CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1344 VNET_DEFINE_STATIC(int, soft_pad);
1345 #define V_soft_pad VNET(soft_pad)
1346 SYSCTL_INT(_net_link_vlan, OID_AUTO, soft_pad, CTLFLAG_RW | CTLFLAG_VNET,
1347 &VNET_NAME(soft_pad), 0,
1348 "pad short frames before tagging");
1351 * For now, make preserving PCP via an mbuf tag optional, as it increases
1352 * per-packet memory allocations and frees. In the future, it would be
1353 * preferable to reuse ether_vtag for this, or similar.
1355 int vlan_mtag_pcp = 0;
1356 SYSCTL_INT(_net_link_vlan, OID_AUTO, mtag_pcp, CTLFLAG_RW,
1358 "Retain VLAN PCP information as packets are passed up the stack");
1361 ether_8021q_frame(struct mbuf **mp, struct ifnet *ife, struct ifnet *p,
1362 struct ether_8021q_tag *qtag)
1367 static const char pad[8]; /* just zeros */
1370 * Pad the frame to the minimum size allowed if told to.
1371 * This option is in accord with IEEE Std 802.1Q, 2003 Ed.,
1372 * paragraph C.4.4.3.b. It can help to work around buggy
1373 * bridges that violate paragraph C.4.4.3.a from the same
1374 * document, i.e., fail to pad short frames after untagging.
1375 * E.g., a tagged frame 66 bytes long (incl. FCS) is OK, but
1376 * untagging it will produce a 62-byte frame, which is a runt
1377 * and requires padding. There are VLAN-enabled network
1378 * devices that just discard such runts instead or mishandle
1381 if (V_soft_pad && p->if_type == IFT_ETHER) {
1382 for (n = ETHERMIN + ETHER_HDR_LEN - (*mp)->m_pkthdr.len;
1383 n > 0; n -= sizeof(pad)) {
1384 if (!m_append(*mp, min(n, sizeof(pad)), pad))
1390 if_printf(ife, "cannot pad short frame");
1396 * If PCP is set in mbuf, use it
1398 if ((*mp)->m_flags & M_VLANTAG) {
1399 qtag->pcp = EVL_PRIOFTAG((*mp)->m_pkthdr.ether_vtag);
1403 * If underlying interface can do VLAN tag insertion itself,
1404 * just pass the packet along. However, we need some way to
1405 * tell the interface where the packet came from so that it
1406 * knows how to find the VLAN tag to use, so we attach a
1407 * packet tag that holds it.
1409 if (vlan_mtag_pcp && (mtag = m_tag_locate(*mp, MTAG_8021Q,
1410 MTAG_8021Q_PCP_OUT, NULL)) != NULL)
1411 tag = EVL_MAKETAG(qtag->vid, *(uint8_t *)(mtag + 1), 0);
1413 tag = EVL_MAKETAG(qtag->vid, qtag->pcp, 0);
1414 if ((p->if_capenable & IFCAP_VLAN_HWTAGGING) &&
1415 (qtag->proto == ETHERTYPE_VLAN)) {
1416 (*mp)->m_pkthdr.ether_vtag = tag;
1417 (*mp)->m_flags |= M_VLANTAG;
1419 *mp = ether_vlanencap_proto(*mp, tag, qtag->proto);
1421 if_printf(ife, "unable to prepend 802.1Q header");
1429 * Allocate an address from the FreeBSD Foundation OUI. This uses a
1430 * cryptographic hash function on the containing jail's name, UUID and the
1431 * interface name to attempt to provide a unique but stable address.
1432 * Pseudo-interfaces which require a MAC address should use this function to
1433 * allocate non-locally-administered addresses.
1436 ether_gen_addr(struct ifnet *ifp, struct ether_addr *hwaddr)
1440 char uuid[HOSTUUIDLEN + 1];
1443 char digest[SHA1_RESULTLEN];
1444 char jailname[MAXHOSTNAMELEN];
1446 getcredhostuuid(curthread->td_ucred, uuid, sizeof(uuid));
1447 if (strncmp(uuid, DEFAULT_HOSTUUID, sizeof(uuid)) == 0) {
1448 /* Fall back to a random mac address. */
1452 /* If each (vnet) jail would also have a unique hostuuid this would not
1454 getjailname(curthread->td_ucred, jailname, sizeof(jailname));
1455 sz = asprintf(&buf, M_TEMP, "%s-%s-%s", uuid, if_name(ifp),
1458 /* Fall back to a random mac address. */
1463 SHA1Update(&ctx, buf, sz);
1464 SHA1Final(digest, &ctx);
1467 addr = ((digest[0] << 16) | (digest[1] << 8) | digest[2]) &
1468 OUI_FREEBSD_GENERATED_MASK;
1469 addr = OUI_FREEBSD(addr);
1470 for (i = 0; i < ETHER_ADDR_LEN; ++i) {
1471 hwaddr->octet[i] = addr >> ((ETHER_ADDR_LEN - i - 1) * 8) &
1477 arc4rand(hwaddr, sizeof(*hwaddr), 0);
1479 hwaddr->octet[0] &= 0xFE;
1480 /* Locally administered. */
1481 hwaddr->octet[0] |= 0x02;
1484 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1485 MODULE_VERSION(ether, 1);