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>
50 #include <sys/module.h>
54 #include <sys/random.h>
55 #include <sys/socket.h>
56 #include <sys/sockio.h>
57 #include <sys/sysctl.h>
63 #include <net/ieee_oui.h>
65 #include <net/if_var.h>
66 #include <net/if_private.h>
67 #include <net/if_arp.h>
68 #include <net/netisr.h>
69 #include <net/route.h>
70 #include <net/if_llc.h>
71 #include <net/if_dl.h>
72 #include <net/if_types.h>
74 #include <net/ethernet.h>
75 #include <net/if_bridgevar.h>
76 #include <net/if_vlan_var.h>
77 #include <net/if_llatbl.h>
79 #include <net/rss_config.h>
82 #include <netpfil/pf/pf_mtag.h>
84 #if defined(INET) || defined(INET6)
85 #include <netinet/in.h>
86 #include <netinet/in_var.h>
87 #include <netinet/if_ether.h>
88 #include <netinet/ip_carp.h>
89 #include <netinet/ip_var.h>
92 #include <netinet6/nd6.h>
94 #include <security/mac/mac_framework.h>
96 #include <crypto/sha1.h>
99 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2);
100 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN);
103 VNET_DEFINE(pfil_head_t, link_pfil_head); /* Packet filter hooks */
105 /* netgraph node hooks for ng_ether(4) */
106 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
107 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
108 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
109 void (*ng_ether_attach_p)(struct ifnet *ifp);
110 void (*ng_ether_detach_p)(struct ifnet *ifp);
112 void (*vlan_input_p)(struct ifnet *, struct mbuf *);
114 /* if_bridge(4) support */
115 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
117 /* if_lagg(4) support */
118 struct mbuf *(*lagg_input_ethernet_p)(struct ifnet *, struct mbuf *);
120 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
121 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
123 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
125 static int ether_requestencap(struct ifnet *, struct if_encap_req *);
127 #define senderr(e) do { error = (e); goto bad;} while (0)
130 update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst)
134 if (src->m_pkthdr.csum_flags & CSUM_IP)
135 csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
136 if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
137 csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
138 if (src->m_pkthdr.csum_flags & CSUM_SCTP)
139 csum_flags |= CSUM_SCTP_VALID;
140 dst->m_pkthdr.csum_flags |= csum_flags;
141 if (csum_flags & CSUM_DATA_VALID)
142 dst->m_pkthdr.csum_data = 0xffff;
146 * Handle link-layer encapsulation requests.
149 ether_requestencap(struct ifnet *ifp, struct if_encap_req *req)
151 struct ether_header *eh;
154 const u_char *lladdr;
156 if (req->rtype != IFENCAP_LL)
159 if (req->bufsize < ETHER_HDR_LEN)
162 eh = (struct ether_header *)req->buf;
163 lladdr = req->lladdr;
166 switch (req->family) {
168 etype = htons(ETHERTYPE_IP);
171 etype = htons(ETHERTYPE_IPV6);
174 ah = (struct arphdr *)req->hdata;
175 ah->ar_hrd = htons(ARPHRD_ETHER);
177 switch(ntohs(ah->ar_op)) {
178 case ARPOP_REVREQUEST:
180 etype = htons(ETHERTYPE_REVARP);
185 etype = htons(ETHERTYPE_ARP);
189 if (req->flags & IFENCAP_FLAG_BROADCAST)
190 lladdr = ifp->if_broadcastaddr;
193 return (EAFNOSUPPORT);
196 memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
197 memcpy(eh->ether_dhost, lladdr, ETHER_ADDR_LEN);
198 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
199 req->bufsize = sizeof(struct ether_header);
205 ether_resolve_addr(struct ifnet *ifp, struct mbuf *m,
206 const struct sockaddr *dst, struct route *ro, u_char *phdr,
207 uint32_t *pflags, struct llentry **plle)
209 uint32_t lleflags = 0;
211 #if defined(INET) || defined(INET6)
212 struct ether_header *eh = (struct ether_header *)phdr;
219 switch (dst->sa_family) {
222 if ((m->m_flags & (M_BCAST | M_MCAST)) == 0)
223 error = arpresolve(ifp, 0, m, dst, phdr, &lleflags,
226 if (m->m_flags & M_BCAST)
227 memcpy(eh->ether_dhost, ifp->if_broadcastaddr,
230 const struct in_addr *a;
231 a = &(((const struct sockaddr_in *)dst)->sin_addr);
232 ETHER_MAP_IP_MULTICAST(a, eh->ether_dhost);
234 etype = htons(ETHERTYPE_IP);
235 memcpy(&eh->ether_type, &etype, sizeof(etype));
236 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
242 if ((m->m_flags & M_MCAST) == 0) {
243 int af = RO_GET_FAMILY(ro, dst);
244 error = nd6_resolve(ifp, LLE_SF(af, 0), m, dst, phdr,
247 const struct in6_addr *a6;
248 a6 = &(((const struct sockaddr_in6 *)dst)->sin6_addr);
249 ETHER_MAP_IPV6_MULTICAST(a6, eh->ether_dhost);
250 etype = htons(ETHERTYPE_IPV6);
251 memcpy(&eh->ether_type, &etype, sizeof(etype));
252 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
257 if_printf(ifp, "can't handle af%d\n", dst->sa_family);
260 return (EAFNOSUPPORT);
263 if (error == EHOSTDOWN) {
264 if (ro != NULL && (ro->ro_flags & RT_HAS_GW) != 0)
265 error = EHOSTUNREACH;
271 *pflags = RT_MAY_LOOP;
272 if (lleflags & LLE_IFADDR)
279 * Ethernet output routine.
280 * Encapsulate a packet of type family for the local net.
281 * Use trailer local net encapsulation if enough data in first
282 * packet leaves a multiple of 512 bytes of data in remainder.
285 ether_output(struct ifnet *ifp, struct mbuf *m,
286 const struct sockaddr *dst, struct route *ro)
289 char linkhdr[ETHER_HDR_LEN], *phdr;
290 struct ether_header *eh;
293 int hlen; /* link layer header length */
295 struct llentry *lle = NULL;
301 /* XXX BPF uses ro_prepend */
302 if (ro->ro_prepend != NULL) {
303 phdr = ro->ro_prepend;
305 } else if (!(m->m_flags & (M_BCAST | M_MCAST))) {
306 if ((ro->ro_flags & RT_LLE_CACHE) != 0) {
309 (lle->la_flags & LLE_VALID) == 0) {
311 lle = NULL; /* redundant */
315 /* if we lookup, keep cache */
319 * Notify LLE code that
323 llentry_provide_feedback(lle);
326 phdr = lle->r_linkdata;
327 hlen = lle->r_hdrlen;
328 pflags = lle->r_flags;
334 error = mac_ifnet_check_transmit(ifp, m);
340 if (ifp->if_flags & IFF_MONITOR)
342 if (!((ifp->if_flags & IFF_UP) &&
343 (ifp->if_drv_flags & IFF_DRV_RUNNING)))
347 /* No prepend data supplied. Try to calculate ourselves. */
349 hlen = ETHER_HDR_LEN;
350 error = ether_resolve_addr(ifp, m, dst, ro, phdr, &pflags,
351 addref ? &lle : NULL);
352 if (addref && lle != NULL)
355 return (error == EWOULDBLOCK ? 0 : error);
358 if ((pflags & RT_L2_ME) != 0) {
359 update_mbuf_csumflags(m, m);
360 return (if_simloop(ifp, m, RO_GET_FAMILY(ro, dst), 0));
362 loop_copy = (pflags & RT_MAY_LOOP) != 0;
365 * Add local net header. If no space in first mbuf,
368 * Note that we do prepend regardless of RT_HAS_HEADER flag.
369 * This is done because BPF code shifts m_data pointer
370 * to the end of ethernet header prior to calling if_output().
372 M_PREPEND(m, hlen, M_NOWAIT);
375 if ((pflags & RT_HAS_HEADER) == 0) {
376 eh = mtod(m, struct ether_header *);
377 memcpy(eh, phdr, hlen);
381 * If a simplex interface, and the packet is being sent to our
382 * Ethernet address or a broadcast address, loopback a copy.
383 * XXX To make a simplex device behave exactly like a duplex
384 * device, we should copy in the case of sending to our own
385 * ethernet address (thus letting the original actually appear
386 * on the wire). However, we don't do that here for security
387 * reasons and compatibility with the original behavior.
389 if ((m->m_flags & M_BCAST) && loop_copy && (ifp->if_flags & IFF_SIMPLEX) &&
390 ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
394 * Because if_simloop() modifies the packet, we need a
395 * writable copy through m_dup() instead of a readonly
396 * one as m_copy[m] would give us. The alternative would
397 * be to modify if_simloop() to handle the readonly mbuf,
398 * but performancewise it is mostly equivalent (trading
399 * extra data copying vs. extra locking).
401 * XXX This is a local workaround. A number of less
402 * often used kernel parts suffer from the same bug.
403 * See PR kern/105943 for a proposed general solution.
405 if ((n = m_dup(m, M_NOWAIT)) != NULL) {
406 update_mbuf_csumflags(m, n);
407 (void)if_simloop(ifp, n, RO_GET_FAMILY(ro, dst), hlen);
409 if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
413 * Bridges require special output handling.
415 if (ifp->if_bridge) {
416 BRIDGE_OUTPUT(ifp, m, error);
420 #if defined(INET) || defined(INET6)
422 (error = (*carp_output_p)(ifp, m, dst)))
426 /* Handle ng_ether(4) processing, if any */
427 if (ifp->if_l2com != NULL) {
428 KASSERT(ng_ether_output_p != NULL,
429 ("ng_ether_output_p is NULL"));
430 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
439 /* Continue with link-layer output */
440 return ether_output_frame(ifp, m);
444 ether_set_pcp(struct mbuf **mp, struct ifnet *ifp, uint8_t pcp)
446 struct ether_8021q_tag qtag;
447 struct ether_header *eh;
449 eh = mtod(*mp, struct ether_header *);
450 if (ntohs(eh->ether_type) == ETHERTYPE_VLAN ||
451 ntohs(eh->ether_type) == ETHERTYPE_QINQ)
456 qtag.proto = ETHERTYPE_VLAN;
457 if (ether_8021q_frame(mp, ifp, ifp, &qtag))
459 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
464 * Ethernet link layer output routine to send a raw frame to the device.
466 * This assumes that the 14 byte Ethernet header is present and contiguous
467 * in the first mbuf (if BRIDGE'ing).
470 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
475 if (pcp != IFNET_PCP_NONE && ifp->if_type != IFT_L2VLAN &&
476 !ether_set_pcp(&m, ifp, pcp))
479 if (PFIL_HOOKED_OUT(V_link_pfil_head))
480 switch (pfil_mbuf_out(V_link_pfil_head, &m, ifp, NULL)) {
488 #if defined(INET6) && defined(INET)
489 /* draft-ietf-6man-ipv6only-flag */
490 /* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */
491 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) {
492 struct ether_header *eh;
494 eh = mtod(m, struct ether_header *);
495 switch (ntohs(eh->ether_type)) {
498 case ETHERTYPE_REVARP:
500 return (EAFNOSUPPORT);
509 * Queue message on interface, update output statistics if successful,
510 * and start output if interface not yet active.
512 * If KMSAN is enabled, use it to verify that the data does not contain
513 * any uninitialized bytes.
515 kmsan_check_mbuf(m, "ether_output");
516 return ((ifp->if_transmit)(ifp, m));
520 * Process a received Ethernet packet; the packet is in the
521 * mbuf chain m with the ethernet header at the front.
524 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
526 struct ether_header *eh;
529 if ((ifp->if_flags & IFF_UP) == 0) {
534 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
535 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
540 if (m->m_len < ETHER_HDR_LEN) {
541 /* XXX maybe should pullup? */
542 if_printf(ifp, "discard frame w/o leading ethernet "
543 "header (len %u pkt len %u)\n",
544 m->m_len, m->m_pkthdr.len);
545 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
549 eh = mtod(m, struct ether_header *);
550 etype = ntohs(eh->ether_type);
551 random_harvest_queue_ether(m, sizeof(*m));
554 #if defined(INET6) && defined(INET)
555 /* draft-ietf-6man-ipv6only-flag */
556 /* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */
557 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) {
561 case ETHERTYPE_REVARP:
571 CURVNET_SET_QUIET(ifp->if_vnet);
573 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
574 if (ETHER_IS_BROADCAST(eh->ether_dhost))
575 m->m_flags |= M_BCAST;
577 m->m_flags |= M_MCAST;
578 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
583 * Tag the mbuf with an appropriate MAC label before any other
584 * consumers can get to it.
586 mac_ifnet_create_mbuf(ifp, m);
590 * Give bpf a chance at the packet.
592 ETHER_BPF_MTAP(ifp, m);
595 * If the CRC is still on the packet, trim it off. We do this once
596 * and once only in case we are re-entered. Nothing else on the
597 * Ethernet receive path expects to see the FCS.
599 if (m->m_flags & M_HASFCS) {
600 m_adj(m, -ETHER_CRC_LEN);
601 m->m_flags &= ~M_HASFCS;
604 if (!(ifp->if_capenable & IFCAP_HWSTATS))
605 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
607 /* Allow monitor mode to claim this frame, after stats are updated. */
608 if (ifp->if_flags & IFF_MONITOR) {
614 /* Handle input from a lagg(4) port */
615 if (ifp->if_type == IFT_IEEE8023ADLAG) {
616 KASSERT(lagg_input_ethernet_p != NULL,
617 ("%s: if_lagg not loaded!", __func__));
618 m = (*lagg_input_ethernet_p)(ifp, m);
620 ifp = m->m_pkthdr.rcvif;
628 * If the hardware did not process an 802.1Q tag, do this now,
629 * to allow 802.1P priority frames to be passed to the main input
632 if ((m->m_flags & M_VLANTAG) == 0 &&
633 ((etype == ETHERTYPE_VLAN) || (etype == ETHERTYPE_QINQ))) {
634 struct ether_vlan_header *evl;
636 if (m->m_len < sizeof(*evl) &&
637 (m = m_pullup(m, sizeof(*evl))) == NULL) {
639 if_printf(ifp, "cannot pullup VLAN header\n");
641 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
646 evl = mtod(m, struct ether_vlan_header *);
647 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
648 m->m_flags |= M_VLANTAG;
650 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
651 ETHER_HDR_LEN - ETHER_TYPE_LEN);
652 m_adj(m, ETHER_VLAN_ENCAP_LEN);
653 eh = mtod(m, struct ether_header *);
656 M_SETFIB(m, ifp->if_fib);
658 /* Allow ng_ether(4) to claim this frame. */
659 if (ifp->if_l2com != NULL) {
660 KASSERT(ng_ether_input_p != NULL,
661 ("%s: ng_ether_input_p is NULL", __func__));
662 m->m_flags &= ~M_PROMISC;
663 (*ng_ether_input_p)(ifp, &m);
668 eh = mtod(m, struct ether_header *);
672 * Allow if_bridge(4) to claim this frame.
674 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
675 * and the frame should be delivered locally.
677 * If M_BRIDGE_INJECT is set, the packet was received directly by the
678 * bridge via netmap, so "ifp" is the bridge itself and the packet
679 * should be re-examined.
681 if (ifp->if_bridge != NULL || (m->m_flags & M_BRIDGE_INJECT) != 0) {
682 m->m_flags &= ~M_PROMISC;
683 BRIDGE_INPUT(ifp, m);
688 eh = mtod(m, struct ether_header *);
691 #if defined(INET) || defined(INET6)
693 * Clear M_PROMISC on frame so that carp(4) will see it when the
694 * mbuf flows up to Layer 3.
695 * FreeBSD's implementation of carp(4) uses the inprotosw
696 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
697 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
698 * is outside the scope of the M_PROMISC test below.
699 * TODO: Maintain a hash table of ethernet addresses other than
700 * ether_dhost which may be active on this ifp.
702 if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
703 m->m_flags &= ~M_PROMISC;
708 * If the frame received was not for our MAC address, set the
709 * M_PROMISC flag on the mbuf chain. The frame may need to
710 * be seen by the rest of the Ethernet input path in case of
711 * re-entry (e.g. bridge, vlan, netgraph) but should not be
712 * seen by upper protocol layers.
714 if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
715 bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
716 m->m_flags |= M_PROMISC;
724 * Ethernet input dispatch; by default, direct dispatch here regardless of
725 * global configuration. However, if RSS is enabled, hook up RSS affinity
726 * so that when deferred or hybrid dispatch is enabled, we can redistribute
729 * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or
730 * not it had already done work distribution via multi-queue. Then we could
731 * direct dispatch in the event load balancing was already complete and
732 * handle the case of interfaces with different capabilities better.
734 * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions
735 * at multiple layers?
737 * XXXRW: For now, enable all this only if RSS is compiled in, although it
738 * works fine without RSS. Need to characterise the performance overhead
739 * of the detour through the netisr code in the event the result is always
743 ether_nh_input(struct mbuf *m)
747 KASSERT(m->m_pkthdr.rcvif != NULL,
748 ("%s: NULL interface pointer", __func__));
749 ether_input_internal(m->m_pkthdr.rcvif, m);
752 static struct netisr_handler ether_nh = {
754 .nh_handler = ether_nh_input,
755 .nh_proto = NETISR_ETHER,
757 .nh_policy = NETISR_POLICY_CPU,
758 .nh_dispatch = NETISR_DISPATCH_DIRECT,
759 .nh_m2cpuid = rss_m2cpuid,
761 .nh_policy = NETISR_POLICY_SOURCE,
762 .nh_dispatch = NETISR_DISPATCH_DIRECT,
767 ether_init(__unused void *arg)
770 netisr_register(ðer_nh);
772 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
775 vnet_ether_init(__unused void *arg)
777 struct pfil_head_args args;
779 args.pa_version = PFIL_VERSION;
780 args.pa_flags = PFIL_IN | PFIL_OUT;
781 args.pa_type = PFIL_TYPE_ETHERNET;
782 args.pa_headname = PFIL_ETHER_NAME;
783 V_link_pfil_head = pfil_head_register(&args);
786 netisr_register_vnet(ðer_nh);
789 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY,
790 vnet_ether_init, NULL);
794 vnet_ether_pfil_destroy(__unused void *arg)
797 pfil_head_unregister(V_link_pfil_head);
799 VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY,
800 vnet_ether_pfil_destroy, NULL);
803 vnet_ether_destroy(__unused void *arg)
806 netisr_unregister_vnet(ðer_nh);
808 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY,
809 vnet_ether_destroy, NULL);
813 ether_input(struct ifnet *ifp, struct mbuf *m)
815 struct epoch_tracker et;
819 needs_epoch = (ifp->if_flags & IFF_NEEDSEPOCH);
822 * This temporary code is here to prevent epoch unaware and unmarked
823 * drivers to panic the system. Once all drivers are taken care of,
824 * the whole INVARIANTS block should go away.
826 if (!needs_epoch && !in_epoch(net_epoch_preempt)) {
827 static bool printedonce;
832 if_printf(ifp, "called %s w/o net epoch! "
833 "PLEASE file a bug report.", __func__);
842 * The drivers are allowed to pass in a chain of packets linked with
843 * m_nextpkt. We split them up into separate packets here and pass
844 * them up. This allows the drivers to amortize the receive lock.
846 CURVNET_SET_QUIET(ifp->if_vnet);
847 if (__predict_false(needs_epoch))
854 * We will rely on rcvif being set properly in the deferred
855 * context, so assert it is correct here.
857 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
858 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p "
859 "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp));
860 netisr_dispatch(NETISR_ETHER, m);
863 if (__predict_false(needs_epoch))
869 * Upper layer processing for a received Ethernet packet.
872 ether_demux(struct ifnet *ifp, struct mbuf *m)
874 struct ether_header *eh;
879 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
881 /* Do not grab PROMISC frames in case we are re-entered. */
882 if (PFIL_HOOKED_IN(V_link_pfil_head) && !(m->m_flags & M_PROMISC)) {
883 i = pfil_mbuf_in(V_link_pfil_head, &m, ifp, NULL);
884 if (i != 0 || m == NULL)
888 eh = mtod(m, struct ether_header *);
889 ether_type = ntohs(eh->ether_type);
892 * If this frame has a VLAN tag other than 0, call vlan_input()
893 * if its module is loaded. Otherwise, drop.
895 if ((m->m_flags & M_VLANTAG) &&
896 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
897 if (ifp->if_vlantrunk == NULL) {
898 if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
902 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
904 /* Clear before possibly re-entering ether_input(). */
905 m->m_flags &= ~M_PROMISC;
906 (*vlan_input_p)(ifp, m);
911 * Pass promiscuously received frames to the upper layer if the user
912 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
914 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
920 * Reset layer specific mbuf flags to avoid confusing upper layers.
922 m->m_flags &= ~M_VLANTAG;
926 * Dispatch frame to upper layer.
928 switch (ether_type) {
935 if (ifp->if_flags & IFF_NOARP) {
936 /* Discard packet if ARP is disabled on interface */
952 /* Strip off Ethernet header. */
953 m_adj(m, ETHER_HDR_LEN);
955 netisr_dispatch(isr, m);
960 * Packet is to be discarded. If netgraph is present,
961 * hand the packet to it for last chance processing;
962 * otherwise dispose of it.
964 if (ifp->if_l2com != NULL) {
965 KASSERT(ng_ether_input_orphan_p != NULL,
966 ("ng_ether_input_orphan_p is NULL"));
967 (*ng_ether_input_orphan_p)(ifp, m);
974 * Convert Ethernet address to printable (loggable) representation.
975 * This routine is for compatibility; it's better to just use
977 * printf("%6D", <pointer to address>, ":");
979 * since there's no static buffer involved.
982 ether_sprintf(const u_char *ap)
984 static char etherbuf[18];
985 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
990 * Perform common duties while attaching to interface list
993 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
997 struct sockaddr_dl *sdl;
999 ifp->if_addrlen = ETHER_ADDR_LEN;
1000 ifp->if_hdrlen = ETHER_HDR_LEN;
1001 ifp->if_mtu = ETHERMTU;
1003 ifp->if_output = ether_output;
1004 ifp->if_input = ether_input;
1005 ifp->if_resolvemulti = ether_resolvemulti;
1006 ifp->if_requestencap = ether_requestencap;
1008 ifp->if_reassign = ether_reassign;
1010 if (ifp->if_baudrate == 0)
1011 ifp->if_baudrate = IF_Mbps(10); /* just a default */
1012 ifp->if_broadcastaddr = etherbroadcastaddr;
1015 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
1016 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1017 sdl->sdl_type = IFT_ETHER;
1018 sdl->sdl_alen = ifp->if_addrlen;
1019 bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
1021 if (ifp->if_hw_addr != NULL)
1022 bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen);
1024 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
1025 if (ng_ether_attach_p != NULL)
1026 (*ng_ether_attach_p)(ifp);
1028 /* Announce Ethernet MAC address if non-zero. */
1029 for (i = 0; i < ifp->if_addrlen; i++)
1032 if (i != ifp->if_addrlen)
1033 if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
1035 uuid_ether_add(LLADDR(sdl));
1037 /* Add necessary bits are setup; announce it now. */
1038 EVENTHANDLER_INVOKE(ether_ifattach_event, ifp);
1039 if (IS_DEFAULT_VNET(curvnet))
1040 devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL);
1044 * Perform common duties while detaching an Ethernet interface
1047 ether_ifdetach(struct ifnet *ifp)
1049 struct sockaddr_dl *sdl;
1051 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr);
1052 uuid_ether_del(LLADDR(sdl));
1054 if (ifp->if_l2com != NULL) {
1055 KASSERT(ng_ether_detach_p != NULL,
1056 ("ng_ether_detach_p is NULL"));
1057 (*ng_ether_detach_p)(ifp);
1066 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused)
1069 if (ifp->if_l2com != NULL) {
1070 KASSERT(ng_ether_detach_p != NULL,
1071 ("ng_ether_detach_p is NULL"));
1072 (*ng_ether_detach_p)(ifp);
1075 if (ng_ether_attach_p != NULL) {
1076 CURVNET_SET_QUIET(new_vnet);
1077 (*ng_ether_attach_p)(ifp);
1083 SYSCTL_DECL(_net_link);
1084 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1089 * This is for reference. We have a table-driven version
1090 * of the little-endian crc32 generator, which is faster
1091 * than the double-loop.
1094 ether_crc32_le(const uint8_t *buf, size_t len)
1101 crc = 0xffffffff; /* initial value */
1103 for (i = 0; i < len; i++) {
1104 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1105 carry = (crc ^ data) & 1;
1108 crc = (crc ^ ETHER_CRC_POLY_LE);
1116 ether_crc32_le(const uint8_t *buf, size_t len)
1118 static const uint32_t crctab[] = {
1119 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1120 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1121 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1122 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1127 crc = 0xffffffff; /* initial value */
1129 for (i = 0; i < len; i++) {
1131 crc = (crc >> 4) ^ crctab[crc & 0xf];
1132 crc = (crc >> 4) ^ crctab[crc & 0xf];
1140 ether_crc32_be(const uint8_t *buf, size_t len)
1143 uint32_t crc, carry;
1147 crc = 0xffffffff; /* initial value */
1149 for (i = 0; i < len; i++) {
1150 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1151 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1154 crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1162 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1164 struct ifaddr *ifa = (struct ifaddr *) data;
1165 struct ifreq *ifr = (struct ifreq *) data;
1170 ifp->if_flags |= IFF_UP;
1172 switch (ifa->ifa_addr->sa_family) {
1175 ifp->if_init(ifp->if_softc); /* before arpwhohas */
1176 arp_ifinit(ifp, ifa);
1180 ifp->if_init(ifp->if_softc);
1186 bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0],
1192 * Set the interface MTU.
1194 if (ifr->ifr_mtu > ETHERMTU) {
1197 ifp->if_mtu = ifr->ifr_mtu;
1202 error = priv_check(curthread, PRIV_NET_SETLANPCP);
1205 if (ifr->ifr_lan_pcp > 7 &&
1206 ifr->ifr_lan_pcp != IFNET_PCP_NONE) {
1209 ifp->if_pcp = ifr->ifr_lan_pcp;
1210 /* broadcast event about PCP change */
1211 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP);
1216 ifr->ifr_lan_pcp = ifp->if_pcp;
1220 error = EINVAL; /* XXX netbsd has ENOTTY??? */
1227 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1228 struct sockaddr *sa)
1230 struct sockaddr_dl *sdl;
1232 struct sockaddr_in *sin;
1235 struct sockaddr_in6 *sin6;
1239 switch(sa->sa_family) {
1242 * No mapping needed. Just check that it's a valid MC address.
1244 sdl = (struct sockaddr_dl *)sa;
1245 e_addr = LLADDR(sdl);
1246 if (!ETHER_IS_MULTICAST(e_addr))
1247 return EADDRNOTAVAIL;
1253 sin = (struct sockaddr_in *)sa;
1254 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1255 return EADDRNOTAVAIL;
1256 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1257 sdl->sdl_alen = ETHER_ADDR_LEN;
1258 e_addr = LLADDR(sdl);
1259 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1260 *llsa = (struct sockaddr *)sdl;
1265 sin6 = (struct sockaddr_in6 *)sa;
1266 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1268 * An IP6 address of 0 means listen to all
1269 * of the Ethernet multicast address used for IP6.
1270 * (This is used for multicast routers.)
1272 ifp->if_flags |= IFF_ALLMULTI;
1276 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1277 return EADDRNOTAVAIL;
1278 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1279 sdl->sdl_alen = ETHER_ADDR_LEN;
1280 e_addr = LLADDR(sdl);
1281 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1282 *llsa = (struct sockaddr *)sdl;
1288 * Well, the text isn't quite right, but it's the name
1291 return EAFNOSUPPORT;
1295 static moduledata_t ether_mod = {
1300 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1302 struct ether_vlan_header vlan;
1305 KASSERT((m->m_flags & M_VLANTAG) != 0,
1306 ("%s: vlan information not present", __func__));
1307 KASSERT(m->m_len >= sizeof(struct ether_header),
1308 ("%s: mbuf not large enough for header", __func__));
1309 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1310 vlan.evl_proto = vlan.evl_encap_proto;
1311 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1312 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1313 m->m_len -= sizeof(struct ether_header);
1314 m->m_data += sizeof(struct ether_header);
1316 * If a data link has been supplied by the caller, then we will need to
1317 * re-create a stack allocated mbuf chain with the following structure:
1319 * (1) mbuf #1 will contain the supplied data link
1320 * (2) mbuf #2 will contain the vlan header
1321 * (3) mbuf #3 will contain the original mbuf's packet data
1323 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1327 mv.m_data = (caddr_t)&vlan;
1328 mv.m_len = sizeof(vlan);
1334 bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1335 m->m_len += sizeof(struct ether_header);
1336 m->m_data -= sizeof(struct ether_header);
1340 ether_vlanencap_proto(struct mbuf *m, uint16_t tag, uint16_t proto)
1342 struct ether_vlan_header *evl;
1344 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT);
1347 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1349 if (m->m_len < sizeof(*evl)) {
1350 m = m_pullup(m, sizeof(*evl));
1356 * Transform the Ethernet header into an Ethernet header
1357 * with 802.1Q encapsulation.
1359 evl = mtod(m, struct ether_vlan_header *);
1360 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1361 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1362 evl->evl_encap_proto = htons(proto);
1363 evl->evl_tag = htons(tag);
1368 ether_bpf_mtap_if(struct ifnet *ifp, struct mbuf *m)
1370 if (bpf_peers_present(ifp->if_bpf)) {
1372 if ((m->m_flags & M_VLANTAG) != 0)
1373 ether_vlan_mtap(ifp->if_bpf, m, NULL, 0);
1375 bpf_mtap(ifp->if_bpf, m);
1379 static SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1380 "IEEE 802.1Q VLAN");
1381 static SYSCTL_NODE(_net_link_vlan, PF_LINK, link,
1382 CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1385 VNET_DEFINE_STATIC(int, soft_pad);
1386 #define V_soft_pad VNET(soft_pad)
1387 SYSCTL_INT(_net_link_vlan, OID_AUTO, soft_pad, CTLFLAG_RW | CTLFLAG_VNET,
1388 &VNET_NAME(soft_pad), 0,
1389 "pad short frames before tagging");
1392 * For now, make preserving PCP via an mbuf tag optional, as it increases
1393 * per-packet memory allocations and frees. In the future, it would be
1394 * preferable to reuse ether_vtag for this, or similar.
1396 VNET_DEFINE(int, vlan_mtag_pcp) = 0;
1397 #define V_vlan_mtag_pcp VNET(vlan_mtag_pcp)
1398 SYSCTL_INT(_net_link_vlan, OID_AUTO, mtag_pcp, CTLFLAG_RW | CTLFLAG_VNET,
1399 &VNET_NAME(vlan_mtag_pcp), 0,
1400 "Retain VLAN PCP information as packets are passed up the stack");
1403 ether_8021q_frame(struct mbuf **mp, struct ifnet *ife, struct ifnet *p,
1404 struct ether_8021q_tag *qtag)
1409 static const char pad[8]; /* just zeros */
1412 * Pad the frame to the minimum size allowed if told to.
1413 * This option is in accord with IEEE Std 802.1Q, 2003 Ed.,
1414 * paragraph C.4.4.3.b. It can help to work around buggy
1415 * bridges that violate paragraph C.4.4.3.a from the same
1416 * document, i.e., fail to pad short frames after untagging.
1417 * E.g., a tagged frame 66 bytes long (incl. FCS) is OK, but
1418 * untagging it will produce a 62-byte frame, which is a runt
1419 * and requires padding. There are VLAN-enabled network
1420 * devices that just discard such runts instead or mishandle
1423 if (V_soft_pad && p->if_type == IFT_ETHER) {
1424 for (n = ETHERMIN + ETHER_HDR_LEN - (*mp)->m_pkthdr.len;
1425 n > 0; n -= sizeof(pad)) {
1426 if (!m_append(*mp, min(n, sizeof(pad)), pad))
1432 if_printf(ife, "cannot pad short frame");
1438 * If PCP is set in mbuf, use it
1440 if ((*mp)->m_flags & M_VLANTAG) {
1441 qtag->pcp = EVL_PRIOFTAG((*mp)->m_pkthdr.ether_vtag);
1445 * If underlying interface can do VLAN tag insertion itself,
1446 * just pass the packet along. However, we need some way to
1447 * tell the interface where the packet came from so that it
1448 * knows how to find the VLAN tag to use, so we attach a
1449 * packet tag that holds it.
1451 if (V_vlan_mtag_pcp && (mtag = m_tag_locate(*mp, MTAG_8021Q,
1452 MTAG_8021Q_PCP_OUT, NULL)) != NULL)
1453 tag = EVL_MAKETAG(qtag->vid, *(uint8_t *)(mtag + 1), 0);
1455 tag = EVL_MAKETAG(qtag->vid, qtag->pcp, 0);
1456 if ((p->if_capenable & IFCAP_VLAN_HWTAGGING) &&
1457 (qtag->proto == ETHERTYPE_VLAN)) {
1458 (*mp)->m_pkthdr.ether_vtag = tag;
1459 (*mp)->m_flags |= M_VLANTAG;
1461 *mp = ether_vlanencap_proto(*mp, tag, qtag->proto);
1463 if_printf(ife, "unable to prepend 802.1Q header");
1471 * Allocate an address from the FreeBSD Foundation OUI. This uses a
1472 * cryptographic hash function on the containing jail's name, UUID and the
1473 * interface name to attempt to provide a unique but stable address.
1474 * Pseudo-interfaces which require a MAC address should use this function to
1475 * allocate non-locally-administered addresses.
1478 ether_gen_addr(struct ifnet *ifp, struct ether_addr *hwaddr)
1482 char uuid[HOSTUUIDLEN + 1];
1485 char digest[SHA1_RESULTLEN];
1486 char jailname[MAXHOSTNAMELEN];
1488 getcredhostuuid(curthread->td_ucred, uuid, sizeof(uuid));
1489 if (strncmp(uuid, DEFAULT_HOSTUUID, sizeof(uuid)) == 0) {
1490 /* Fall back to a random mac address. */
1494 /* If each (vnet) jail would also have a unique hostuuid this would not
1496 getjailname(curthread->td_ucred, jailname, sizeof(jailname));
1497 sz = asprintf(&buf, M_TEMP, "%s-%s-%s", uuid, if_name(ifp),
1500 /* Fall back to a random mac address. */
1505 SHA1Update(&ctx, buf, sz);
1506 SHA1Final(digest, &ctx);
1509 addr = ((digest[0] << 16) | (digest[1] << 8) | digest[2]) &
1510 OUI_FREEBSD_GENERATED_MASK;
1511 addr = OUI_FREEBSD(addr);
1512 for (i = 0; i < ETHER_ADDR_LEN; ++i) {
1513 hwaddr->octet[i] = addr >> ((ETHER_ADDR_LEN - i - 1) * 8) &
1519 arc4rand(hwaddr, sizeof(*hwaddr), 0);
1521 hwaddr->octet[0] &= 0xFE;
1522 /* Locally administered. */
1523 hwaddr->octet[0] |= 0x02;
1526 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1527 MODULE_VERSION(ether, 1);