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>
60 #include <net/ieee_oui.h>
62 #include <net/if_var.h>
63 #include <net/if_arp.h>
64 #include <net/netisr.h>
65 #include <net/route.h>
66 #include <net/if_llc.h>
67 #include <net/if_dl.h>
68 #include <net/if_types.h>
70 #include <net/ethernet.h>
71 #include <net/if_bridgevar.h>
72 #include <net/if_vlan_var.h>
73 #include <net/if_llatbl.h>
75 #include <net/rss_config.h>
78 #include <netpfil/pf/pf_mtag.h>
80 #if defined(INET) || defined(INET6)
81 #include <netinet/in.h>
82 #include <netinet/in_var.h>
83 #include <netinet/if_ether.h>
84 #include <netinet/ip_carp.h>
85 #include <netinet/ip_var.h>
88 #include <netinet6/nd6.h>
90 #include <security/mac/mac_framework.h>
92 #include <crypto/sha1.h>
95 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2);
96 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN);
99 VNET_DEFINE(pfil_head_t, link_pfil_head); /* Packet filter hooks */
101 /* netgraph node hooks for ng_ether(4) */
102 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
103 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
104 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
105 void (*ng_ether_attach_p)(struct ifnet *ifp);
106 void (*ng_ether_detach_p)(struct ifnet *ifp);
108 void (*vlan_input_p)(struct ifnet *, struct mbuf *);
110 /* if_bridge(4) support */
111 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
113 /* if_lagg(4) support */
114 struct mbuf *(*lagg_input_ethernet_p)(struct ifnet *, struct mbuf *);
116 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
117 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
119 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
121 static int ether_requestencap(struct ifnet *, struct if_encap_req *);
123 #define senderr(e) do { error = (e); goto bad;} while (0)
126 update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst)
130 if (src->m_pkthdr.csum_flags & CSUM_IP)
131 csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
132 if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
133 csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
134 if (src->m_pkthdr.csum_flags & CSUM_SCTP)
135 csum_flags |= CSUM_SCTP_VALID;
136 dst->m_pkthdr.csum_flags |= csum_flags;
137 if (csum_flags & CSUM_DATA_VALID)
138 dst->m_pkthdr.csum_data = 0xffff;
142 * Handle link-layer encapsulation requests.
145 ether_requestencap(struct ifnet *ifp, struct if_encap_req *req)
147 struct ether_header *eh;
150 const u_char *lladdr;
152 if (req->rtype != IFENCAP_LL)
155 if (req->bufsize < ETHER_HDR_LEN)
158 eh = (struct ether_header *)req->buf;
159 lladdr = req->lladdr;
162 switch (req->family) {
164 etype = htons(ETHERTYPE_IP);
167 etype = htons(ETHERTYPE_IPV6);
170 ah = (struct arphdr *)req->hdata;
171 ah->ar_hrd = htons(ARPHRD_ETHER);
173 switch(ntohs(ah->ar_op)) {
174 case ARPOP_REVREQUEST:
176 etype = htons(ETHERTYPE_REVARP);
181 etype = htons(ETHERTYPE_ARP);
185 if (req->flags & IFENCAP_FLAG_BROADCAST)
186 lladdr = ifp->if_broadcastaddr;
189 return (EAFNOSUPPORT);
192 memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
193 memcpy(eh->ether_dhost, lladdr, ETHER_ADDR_LEN);
194 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
195 req->bufsize = sizeof(struct ether_header);
201 ether_resolve_addr(struct ifnet *ifp, struct mbuf *m,
202 const struct sockaddr *dst, struct route *ro, u_char *phdr,
203 uint32_t *pflags, struct llentry **plle)
205 struct ether_header *eh;
206 uint32_t lleflags = 0;
208 #if defined(INET) || defined(INET6)
214 eh = (struct ether_header *)phdr;
216 switch (dst->sa_family) {
219 if ((m->m_flags & (M_BCAST | M_MCAST)) == 0)
220 error = arpresolve(ifp, 0, m, dst, phdr, &lleflags,
223 if (m->m_flags & M_BCAST)
224 memcpy(eh->ether_dhost, ifp->if_broadcastaddr,
227 const struct in_addr *a;
228 a = &(((const struct sockaddr_in *)dst)->sin_addr);
229 ETHER_MAP_IP_MULTICAST(a, eh->ether_dhost);
231 etype = htons(ETHERTYPE_IP);
232 memcpy(&eh->ether_type, &etype, sizeof(etype));
233 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
239 if ((m->m_flags & M_MCAST) == 0) {
240 int af = RO_GET_FAMILY(ro, dst);
241 error = nd6_resolve(ifp, LLE_SF(af, 0), m, dst, phdr,
244 const struct in6_addr *a6;
245 a6 = &(((const struct sockaddr_in6 *)dst)->sin6_addr);
246 ETHER_MAP_IPV6_MULTICAST(a6, eh->ether_dhost);
247 etype = htons(ETHERTYPE_IPV6);
248 memcpy(&eh->ether_type, &etype, sizeof(etype));
249 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
254 if_printf(ifp, "can't handle af%d\n", dst->sa_family);
257 return (EAFNOSUPPORT);
260 if (error == EHOSTDOWN) {
261 if (ro != NULL && (ro->ro_flags & RT_HAS_GW) != 0)
262 error = EHOSTUNREACH;
268 *pflags = RT_MAY_LOOP;
269 if (lleflags & LLE_IFADDR)
276 * Ethernet output routine.
277 * Encapsulate a packet of type family for the local net.
278 * Use trailer local net encapsulation if enough data in first
279 * packet leaves a multiple of 512 bytes of data in remainder.
282 ether_output(struct ifnet *ifp, struct mbuf *m,
283 const struct sockaddr *dst, struct route *ro)
286 char linkhdr[ETHER_HDR_LEN], *phdr;
287 struct ether_header *eh;
290 int hlen; /* link layer header length */
292 struct llentry *lle = NULL;
298 /* XXX BPF uses ro_prepend */
299 if (ro->ro_prepend != NULL) {
300 phdr = ro->ro_prepend;
302 } else if (!(m->m_flags & (M_BCAST | M_MCAST))) {
303 if ((ro->ro_flags & RT_LLE_CACHE) != 0) {
306 (lle->la_flags & LLE_VALID) == 0) {
308 lle = NULL; /* redundant */
312 /* if we lookup, keep cache */
316 * Notify LLE code that
320 llentry_provide_feedback(lle);
323 phdr = lle->r_linkdata;
324 hlen = lle->r_hdrlen;
325 pflags = lle->r_flags;
331 error = mac_ifnet_check_transmit(ifp, m);
337 if (ifp->if_flags & IFF_MONITOR)
339 if (!((ifp->if_flags & IFF_UP) &&
340 (ifp->if_drv_flags & IFF_DRV_RUNNING)))
344 /* No prepend data supplied. Try to calculate ourselves. */
346 hlen = ETHER_HDR_LEN;
347 error = ether_resolve_addr(ifp, m, dst, ro, phdr, &pflags,
348 addref ? &lle : NULL);
349 if (addref && lle != NULL)
352 return (error == EWOULDBLOCK ? 0 : error);
355 if ((pflags & RT_L2_ME) != 0) {
356 update_mbuf_csumflags(m, m);
357 return (if_simloop(ifp, m, RO_GET_FAMILY(ro, dst), 0));
359 loop_copy = (pflags & RT_MAY_LOOP) != 0;
362 * Add local net header. If no space in first mbuf,
365 * Note that we do prepend regardless of RT_HAS_HEADER flag.
366 * This is done because BPF code shifts m_data pointer
367 * to the end of ethernet header prior to calling if_output().
369 M_PREPEND(m, hlen, M_NOWAIT);
372 if ((pflags & RT_HAS_HEADER) == 0) {
373 eh = mtod(m, struct ether_header *);
374 memcpy(eh, phdr, hlen);
378 * If a simplex interface, and the packet is being sent to our
379 * Ethernet address or a broadcast address, loopback a copy.
380 * XXX To make a simplex device behave exactly like a duplex
381 * device, we should copy in the case of sending to our own
382 * ethernet address (thus letting the original actually appear
383 * on the wire). However, we don't do that here for security
384 * reasons and compatibility with the original behavior.
386 if ((m->m_flags & M_BCAST) && loop_copy && (ifp->if_flags & IFF_SIMPLEX) &&
387 ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
391 * Because if_simloop() modifies the packet, we need a
392 * writable copy through m_dup() instead of a readonly
393 * one as m_copy[m] would give us. The alternative would
394 * be to modify if_simloop() to handle the readonly mbuf,
395 * but performancewise it is mostly equivalent (trading
396 * extra data copying vs. extra locking).
398 * XXX This is a local workaround. A number of less
399 * often used kernel parts suffer from the same bug.
400 * See PR kern/105943 for a proposed general solution.
402 if ((n = m_dup(m, M_NOWAIT)) != NULL) {
403 update_mbuf_csumflags(m, n);
404 (void)if_simloop(ifp, n, RO_GET_FAMILY(ro, dst), hlen);
406 if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
410 * Bridges require special output handling.
412 if (ifp->if_bridge) {
413 BRIDGE_OUTPUT(ifp, m, error);
417 #if defined(INET) || defined(INET6)
419 (error = (*carp_output_p)(ifp, m, dst)))
423 /* Handle ng_ether(4) processing, if any */
424 if (ifp->if_l2com != NULL) {
425 KASSERT(ng_ether_output_p != NULL,
426 ("ng_ether_output_p is NULL"));
427 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
436 /* Continue with link-layer output */
437 return ether_output_frame(ifp, m);
441 ether_set_pcp(struct mbuf **mp, struct ifnet *ifp, uint8_t pcp)
443 struct ether_8021q_tag qtag;
444 struct ether_header *eh;
446 eh = mtod(*mp, struct ether_header *);
447 if (ntohs(eh->ether_type) == ETHERTYPE_VLAN ||
448 ntohs(eh->ether_type) == ETHERTYPE_QINQ)
453 qtag.proto = ETHERTYPE_VLAN;
454 if (ether_8021q_frame(mp, ifp, ifp, &qtag))
456 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
461 * Ethernet link layer output routine to send a raw frame to the device.
463 * This assumes that the 14 byte Ethernet header is present and contiguous
464 * in the first mbuf (if BRIDGE'ing).
467 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
472 if (pcp != IFNET_PCP_NONE && ifp->if_type != IFT_L2VLAN &&
473 !ether_set_pcp(&m, ifp, pcp))
476 if (PFIL_HOOKED_OUT(V_link_pfil_head))
477 switch (pfil_run_hooks(V_link_pfil_head, &m, ifp, PFIL_OUT,
486 #if defined(INET6) && defined(INET)
487 /* draft-ietf-6man-ipv6only-flag */
488 /* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */
489 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) {
490 struct ether_header *eh;
492 eh = mtod(m, struct ether_header *);
493 switch (ntohs(eh->ether_type)) {
496 case ETHERTYPE_REVARP:
498 return (EAFNOSUPPORT);
507 * Queue message on interface, update output statistics if successful,
508 * and start output if interface not yet active.
510 * If KMSAN is enabled, use it to verify that the data does not contain
511 * any uninitialized bytes.
513 kmsan_check_mbuf(m, "ether_output");
514 return ((ifp->if_transmit)(ifp, m));
518 * Process a received Ethernet packet; the packet is in the
519 * mbuf chain m with the ethernet header at the front.
522 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
524 struct ether_header *eh;
527 if ((ifp->if_flags & IFF_UP) == 0) {
532 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
533 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
538 if (m->m_len < ETHER_HDR_LEN) {
539 /* XXX maybe should pullup? */
540 if_printf(ifp, "discard frame w/o leading ethernet "
541 "header (len %u pkt len %u)\n",
542 m->m_len, m->m_pkthdr.len);
543 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
547 eh = mtod(m, struct ether_header *);
548 etype = ntohs(eh->ether_type);
549 random_harvest_queue_ether(m, sizeof(*m));
552 #if defined(INET6) && defined(INET)
553 /* draft-ietf-6man-ipv6only-flag */
554 /* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */
555 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) {
559 case ETHERTYPE_REVARP:
569 CURVNET_SET_QUIET(ifp->if_vnet);
571 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
572 if (ETHER_IS_BROADCAST(eh->ether_dhost))
573 m->m_flags |= M_BCAST;
575 m->m_flags |= M_MCAST;
576 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
581 * Tag the mbuf with an appropriate MAC label before any other
582 * consumers can get to it.
584 mac_ifnet_create_mbuf(ifp, m);
588 * Give bpf a chance at the packet.
590 ETHER_BPF_MTAP(ifp, m);
593 * If the CRC is still on the packet, trim it off. We do this once
594 * and once only in case we are re-entered. Nothing else on the
595 * Ethernet receive path expects to see the FCS.
597 if (m->m_flags & M_HASFCS) {
598 m_adj(m, -ETHER_CRC_LEN);
599 m->m_flags &= ~M_HASFCS;
602 if (!(ifp->if_capenable & IFCAP_HWSTATS))
603 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
605 /* Allow monitor mode to claim this frame, after stats are updated. */
606 if (ifp->if_flags & IFF_MONITOR) {
612 /* Handle input from a lagg(4) port */
613 if (ifp->if_type == IFT_IEEE8023ADLAG) {
614 KASSERT(lagg_input_ethernet_p != NULL,
615 ("%s: if_lagg not loaded!", __func__));
616 m = (*lagg_input_ethernet_p)(ifp, m);
618 ifp = m->m_pkthdr.rcvif;
626 * If the hardware did not process an 802.1Q tag, do this now,
627 * to allow 802.1P priority frames to be passed to the main input
630 if ((m->m_flags & M_VLANTAG) == 0 &&
631 ((etype == ETHERTYPE_VLAN) || (etype == ETHERTYPE_QINQ))) {
632 struct ether_vlan_header *evl;
634 if (m->m_len < sizeof(*evl) &&
635 (m = m_pullup(m, sizeof(*evl))) == NULL) {
637 if_printf(ifp, "cannot pullup VLAN header\n");
639 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
644 evl = mtod(m, struct ether_vlan_header *);
645 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
646 m->m_flags |= M_VLANTAG;
648 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
649 ETHER_HDR_LEN - ETHER_TYPE_LEN);
650 m_adj(m, ETHER_VLAN_ENCAP_LEN);
651 eh = mtod(m, struct ether_header *);
654 M_SETFIB(m, ifp->if_fib);
656 /* Allow ng_ether(4) to claim this frame. */
657 if (ifp->if_l2com != NULL) {
658 KASSERT(ng_ether_input_p != NULL,
659 ("%s: ng_ether_input_p is NULL", __func__));
660 m->m_flags &= ~M_PROMISC;
661 (*ng_ether_input_p)(ifp, &m);
666 eh = mtod(m, struct ether_header *);
670 * Allow if_bridge(4) to claim this frame.
671 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
672 * and the frame should be delivered locally.
674 if (ifp->if_bridge != NULL) {
675 m->m_flags &= ~M_PROMISC;
676 BRIDGE_INPUT(ifp, m);
681 eh = mtod(m, struct ether_header *);
684 #if defined(INET) || defined(INET6)
686 * Clear M_PROMISC on frame so that carp(4) will see it when the
687 * mbuf flows up to Layer 3.
688 * FreeBSD's implementation of carp(4) uses the inprotosw
689 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
690 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
691 * is outside the scope of the M_PROMISC test below.
692 * TODO: Maintain a hash table of ethernet addresses other than
693 * ether_dhost which may be active on this ifp.
695 if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
696 m->m_flags &= ~M_PROMISC;
701 * If the frame received was not for our MAC address, set the
702 * M_PROMISC flag on the mbuf chain. The frame may need to
703 * be seen by the rest of the Ethernet input path in case of
704 * re-entry (e.g. bridge, vlan, netgraph) but should not be
705 * seen by upper protocol layers.
707 if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
708 bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
709 m->m_flags |= M_PROMISC;
717 * Ethernet input dispatch; by default, direct dispatch here regardless of
718 * global configuration. However, if RSS is enabled, hook up RSS affinity
719 * so that when deferred or hybrid dispatch is enabled, we can redistribute
722 * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or
723 * not it had already done work distribution via multi-queue. Then we could
724 * direct dispatch in the event load balancing was already complete and
725 * handle the case of interfaces with different capabilities better.
727 * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions
728 * at multiple layers?
730 * XXXRW: For now, enable all this only if RSS is compiled in, although it
731 * works fine without RSS. Need to characterise the performance overhead
732 * of the detour through the netisr code in the event the result is always
736 ether_nh_input(struct mbuf *m)
740 KASSERT(m->m_pkthdr.rcvif != NULL,
741 ("%s: NULL interface pointer", __func__));
742 ether_input_internal(m->m_pkthdr.rcvif, m);
745 static struct netisr_handler ether_nh = {
747 .nh_handler = ether_nh_input,
748 .nh_proto = NETISR_ETHER,
750 .nh_policy = NETISR_POLICY_CPU,
751 .nh_dispatch = NETISR_DISPATCH_DIRECT,
752 .nh_m2cpuid = rss_m2cpuid,
754 .nh_policy = NETISR_POLICY_SOURCE,
755 .nh_dispatch = NETISR_DISPATCH_DIRECT,
760 ether_init(__unused void *arg)
763 netisr_register(ðer_nh);
765 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
768 vnet_ether_init(__unused void *arg)
770 struct pfil_head_args args;
772 args.pa_version = PFIL_VERSION;
773 args.pa_flags = PFIL_IN | PFIL_OUT;
774 args.pa_type = PFIL_TYPE_ETHERNET;
775 args.pa_headname = PFIL_ETHER_NAME;
776 V_link_pfil_head = pfil_head_register(&args);
779 netisr_register_vnet(ðer_nh);
782 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY,
783 vnet_ether_init, NULL);
787 vnet_ether_pfil_destroy(__unused void *arg)
790 pfil_head_unregister(V_link_pfil_head);
792 VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY,
793 vnet_ether_pfil_destroy, NULL);
796 vnet_ether_destroy(__unused void *arg)
799 netisr_unregister_vnet(ðer_nh);
801 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY,
802 vnet_ether_destroy, NULL);
806 ether_input(struct ifnet *ifp, struct mbuf *m)
808 struct epoch_tracker et;
812 needs_epoch = !(ifp->if_flags & IFF_KNOWSEPOCH);
815 * The drivers are allowed to pass in a chain of packets linked with
816 * m_nextpkt. We split them up into separate packets here and pass
817 * them up. This allows the drivers to amortize the receive lock.
819 CURVNET_SET_QUIET(ifp->if_vnet);
820 if (__predict_false(needs_epoch))
827 * We will rely on rcvif being set properly in the deferred
828 * context, so assert it is correct here.
830 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
831 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p "
832 "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp));
833 netisr_dispatch(NETISR_ETHER, m);
836 if (__predict_false(needs_epoch))
842 * Upper layer processing for a received Ethernet packet.
845 ether_demux(struct ifnet *ifp, struct mbuf *m)
847 struct ether_header *eh;
852 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
854 /* Do not grab PROMISC frames in case we are re-entered. */
855 if (PFIL_HOOKED_IN(V_link_pfil_head) && !(m->m_flags & M_PROMISC)) {
856 i = pfil_run_hooks(V_link_pfil_head, &m, ifp, PFIL_IN, NULL);
857 if (i != 0 || m == NULL)
861 eh = mtod(m, struct ether_header *);
862 ether_type = ntohs(eh->ether_type);
865 * If this frame has a VLAN tag other than 0, call vlan_input()
866 * if its module is loaded. Otherwise, drop.
868 if ((m->m_flags & M_VLANTAG) &&
869 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
870 if (ifp->if_vlantrunk == NULL) {
871 if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
875 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
877 /* Clear before possibly re-entering ether_input(). */
878 m->m_flags &= ~M_PROMISC;
879 (*vlan_input_p)(ifp, m);
884 * Pass promiscuously received frames to the upper layer if the user
885 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
887 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
893 * Reset layer specific mbuf flags to avoid confusing upper layers.
894 * Strip off Ethernet header.
896 m->m_flags &= ~M_VLANTAG;
898 m_adj(m, ETHER_HDR_LEN);
901 * Dispatch frame to upper layer.
903 switch (ether_type) {
910 if (ifp->if_flags & IFF_NOARP) {
911 /* Discard packet if ARP is disabled on interface */
926 netisr_dispatch(isr, m);
931 * Packet is to be discarded. If netgraph is present,
932 * hand the packet to it for last chance processing;
933 * otherwise dispose of it.
935 if (ifp->if_l2com != NULL) {
936 KASSERT(ng_ether_input_orphan_p != NULL,
937 ("ng_ether_input_orphan_p is NULL"));
939 * Put back the ethernet header so netgraph has a
940 * consistent view of inbound packets.
942 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
943 (*ng_ether_input_orphan_p)(ifp, m);
950 * Convert Ethernet address to printable (loggable) representation.
951 * This routine is for compatibility; it's better to just use
953 * printf("%6D", <pointer to address>, ":");
955 * since there's no static buffer involved.
958 ether_sprintf(const u_char *ap)
960 static char etherbuf[18];
961 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
966 * Perform common duties while attaching to interface list
969 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
973 struct sockaddr_dl *sdl;
975 ifp->if_addrlen = ETHER_ADDR_LEN;
976 ifp->if_hdrlen = ETHER_HDR_LEN;
977 ifp->if_mtu = ETHERMTU;
979 ifp->if_output = ether_output;
980 ifp->if_input = ether_input;
981 ifp->if_resolvemulti = ether_resolvemulti;
982 ifp->if_requestencap = ether_requestencap;
984 ifp->if_reassign = ether_reassign;
986 if (ifp->if_baudrate == 0)
987 ifp->if_baudrate = IF_Mbps(10); /* just a default */
988 ifp->if_broadcastaddr = etherbroadcastaddr;
991 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
992 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
993 sdl->sdl_type = IFT_ETHER;
994 sdl->sdl_alen = ifp->if_addrlen;
995 bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
997 if (ifp->if_hw_addr != NULL)
998 bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen);
1000 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
1001 if (ng_ether_attach_p != NULL)
1002 (*ng_ether_attach_p)(ifp);
1004 /* Announce Ethernet MAC address if non-zero. */
1005 for (i = 0; i < ifp->if_addrlen; i++)
1008 if (i != ifp->if_addrlen)
1009 if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
1011 uuid_ether_add(LLADDR(sdl));
1013 /* Add necessary bits are setup; announce it now. */
1014 EVENTHANDLER_INVOKE(ether_ifattach_event, ifp);
1015 if (IS_DEFAULT_VNET(curvnet))
1016 devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL);
1020 * Perform common duties while detaching an Ethernet interface
1023 ether_ifdetach(struct ifnet *ifp)
1025 struct sockaddr_dl *sdl;
1027 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr);
1028 uuid_ether_del(LLADDR(sdl));
1030 if (ifp->if_l2com != NULL) {
1031 KASSERT(ng_ether_detach_p != NULL,
1032 ("ng_ether_detach_p is NULL"));
1033 (*ng_ether_detach_p)(ifp);
1042 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused)
1045 if (ifp->if_l2com != NULL) {
1046 KASSERT(ng_ether_detach_p != NULL,
1047 ("ng_ether_detach_p is NULL"));
1048 (*ng_ether_detach_p)(ifp);
1051 if (ng_ether_attach_p != NULL) {
1052 CURVNET_SET_QUIET(new_vnet);
1053 (*ng_ether_attach_p)(ifp);
1059 SYSCTL_DECL(_net_link);
1060 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1065 * This is for reference. We have a table-driven version
1066 * of the little-endian crc32 generator, which is faster
1067 * than the double-loop.
1070 ether_crc32_le(const uint8_t *buf, size_t len)
1077 crc = 0xffffffff; /* initial value */
1079 for (i = 0; i < len; i++) {
1080 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1081 carry = (crc ^ data) & 1;
1084 crc = (crc ^ ETHER_CRC_POLY_LE);
1092 ether_crc32_le(const uint8_t *buf, size_t len)
1094 static const uint32_t crctab[] = {
1095 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1096 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1097 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1098 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1103 crc = 0xffffffff; /* initial value */
1105 for (i = 0; i < len; i++) {
1107 crc = (crc >> 4) ^ crctab[crc & 0xf];
1108 crc = (crc >> 4) ^ crctab[crc & 0xf];
1116 ether_crc32_be(const uint8_t *buf, size_t len)
1119 uint32_t crc, carry;
1123 crc = 0xffffffff; /* initial value */
1125 for (i = 0; i < len; i++) {
1126 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1127 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1130 crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1138 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1140 struct ifaddr *ifa = (struct ifaddr *) data;
1141 struct ifreq *ifr = (struct ifreq *) data;
1146 ifp->if_flags |= IFF_UP;
1148 switch (ifa->ifa_addr->sa_family) {
1151 ifp->if_init(ifp->if_softc); /* before arpwhohas */
1152 arp_ifinit(ifp, ifa);
1156 ifp->if_init(ifp->if_softc);
1162 bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0],
1168 * Set the interface MTU.
1170 if (ifr->ifr_mtu > ETHERMTU) {
1173 ifp->if_mtu = ifr->ifr_mtu;
1178 error = priv_check(curthread, PRIV_NET_SETLANPCP);
1181 if (ifr->ifr_lan_pcp > 7 &&
1182 ifr->ifr_lan_pcp != IFNET_PCP_NONE) {
1185 ifp->if_pcp = ifr->ifr_lan_pcp;
1186 /* broadcast event about PCP change */
1187 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP);
1192 ifr->ifr_lan_pcp = ifp->if_pcp;
1196 error = EINVAL; /* XXX netbsd has ENOTTY??? */
1203 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1204 struct sockaddr *sa)
1206 struct sockaddr_dl *sdl;
1208 struct sockaddr_in *sin;
1211 struct sockaddr_in6 *sin6;
1215 switch(sa->sa_family) {
1218 * No mapping needed. Just check that it's a valid MC address.
1220 sdl = (struct sockaddr_dl *)sa;
1221 e_addr = LLADDR(sdl);
1222 if (!ETHER_IS_MULTICAST(e_addr))
1223 return EADDRNOTAVAIL;
1229 sin = (struct sockaddr_in *)sa;
1230 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1231 return EADDRNOTAVAIL;
1232 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1233 sdl->sdl_alen = ETHER_ADDR_LEN;
1234 e_addr = LLADDR(sdl);
1235 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1236 *llsa = (struct sockaddr *)sdl;
1241 sin6 = (struct sockaddr_in6 *)sa;
1242 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1244 * An IP6 address of 0 means listen to all
1245 * of the Ethernet multicast address used for IP6.
1246 * (This is used for multicast routers.)
1248 ifp->if_flags |= IFF_ALLMULTI;
1252 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1253 return EADDRNOTAVAIL;
1254 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1255 sdl->sdl_alen = ETHER_ADDR_LEN;
1256 e_addr = LLADDR(sdl);
1257 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1258 *llsa = (struct sockaddr *)sdl;
1264 * Well, the text isn't quite right, but it's the name
1267 return EAFNOSUPPORT;
1271 static moduledata_t ether_mod = {
1276 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1278 struct ether_vlan_header vlan;
1281 KASSERT((m->m_flags & M_VLANTAG) != 0,
1282 ("%s: vlan information not present", __func__));
1283 KASSERT(m->m_len >= sizeof(struct ether_header),
1284 ("%s: mbuf not large enough for header", __func__));
1285 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1286 vlan.evl_proto = vlan.evl_encap_proto;
1287 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1288 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1289 m->m_len -= sizeof(struct ether_header);
1290 m->m_data += sizeof(struct ether_header);
1292 * If a data link has been supplied by the caller, then we will need to
1293 * re-create a stack allocated mbuf chain with the following structure:
1295 * (1) mbuf #1 will contain the supplied data link
1296 * (2) mbuf #2 will contain the vlan header
1297 * (3) mbuf #3 will contain the original mbuf's packet data
1299 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1303 mv.m_data = (caddr_t)&vlan;
1304 mv.m_len = sizeof(vlan);
1310 bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1311 m->m_len += sizeof(struct ether_header);
1312 m->m_data -= sizeof(struct ether_header);
1316 ether_vlanencap_proto(struct mbuf *m, uint16_t tag, uint16_t proto)
1318 struct ether_vlan_header *evl;
1320 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT);
1323 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1325 if (m->m_len < sizeof(*evl)) {
1326 m = m_pullup(m, sizeof(*evl));
1332 * Transform the Ethernet header into an Ethernet header
1333 * with 802.1Q encapsulation.
1335 evl = mtod(m, struct ether_vlan_header *);
1336 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1337 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1338 evl->evl_encap_proto = htons(proto);
1339 evl->evl_tag = htons(tag);
1343 static SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1344 "IEEE 802.1Q VLAN");
1345 static SYSCTL_NODE(_net_link_vlan, PF_LINK, link,
1346 CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1349 VNET_DEFINE_STATIC(int, soft_pad);
1350 #define V_soft_pad VNET(soft_pad)
1351 SYSCTL_INT(_net_link_vlan, OID_AUTO, soft_pad, CTLFLAG_RW | CTLFLAG_VNET,
1352 &VNET_NAME(soft_pad), 0,
1353 "pad short frames before tagging");
1356 * For now, make preserving PCP via an mbuf tag optional, as it increases
1357 * per-packet memory allocations and frees. In the future, it would be
1358 * preferable to reuse ether_vtag for this, or similar.
1360 VNET_DEFINE(int, vlan_mtag_pcp) = 0;
1361 #define V_vlan_mtag_pcp VNET(vlan_mtag_pcp)
1362 SYSCTL_INT(_net_link_vlan, OID_AUTO, mtag_pcp, CTLFLAG_RW | CTLFLAG_VNET,
1363 &VNET_NAME(vlan_mtag_pcp), 0,
1364 "Retain VLAN PCP information as packets are passed up the stack");
1367 ether_8021q_frame(struct mbuf **mp, struct ifnet *ife, struct ifnet *p,
1368 struct ether_8021q_tag *qtag)
1373 static const char pad[8]; /* just zeros */
1376 * Pad the frame to the minimum size allowed if told to.
1377 * This option is in accord with IEEE Std 802.1Q, 2003 Ed.,
1378 * paragraph C.4.4.3.b. It can help to work around buggy
1379 * bridges that violate paragraph C.4.4.3.a from the same
1380 * document, i.e., fail to pad short frames after untagging.
1381 * E.g., a tagged frame 66 bytes long (incl. FCS) is OK, but
1382 * untagging it will produce a 62-byte frame, which is a runt
1383 * and requires padding. There are VLAN-enabled network
1384 * devices that just discard such runts instead or mishandle
1387 if (V_soft_pad && p->if_type == IFT_ETHER) {
1388 for (n = ETHERMIN + ETHER_HDR_LEN - (*mp)->m_pkthdr.len;
1389 n > 0; n -= sizeof(pad)) {
1390 if (!m_append(*mp, min(n, sizeof(pad)), pad))
1396 if_printf(ife, "cannot pad short frame");
1402 * If PCP is set in mbuf, use it
1404 if ((*mp)->m_flags & M_VLANTAG) {
1405 qtag->pcp = EVL_PRIOFTAG((*mp)->m_pkthdr.ether_vtag);
1409 * If underlying interface can do VLAN tag insertion itself,
1410 * just pass the packet along. However, we need some way to
1411 * tell the interface where the packet came from so that it
1412 * knows how to find the VLAN tag to use, so we attach a
1413 * packet tag that holds it.
1415 if (V_vlan_mtag_pcp && (mtag = m_tag_locate(*mp, MTAG_8021Q,
1416 MTAG_8021Q_PCP_OUT, NULL)) != NULL)
1417 tag = EVL_MAKETAG(qtag->vid, *(uint8_t *)(mtag + 1), 0);
1419 tag = EVL_MAKETAG(qtag->vid, qtag->pcp, 0);
1420 if ((p->if_capenable & IFCAP_VLAN_HWTAGGING) &&
1421 (qtag->proto == ETHERTYPE_VLAN)) {
1422 (*mp)->m_pkthdr.ether_vtag = tag;
1423 (*mp)->m_flags |= M_VLANTAG;
1425 *mp = ether_vlanencap_proto(*mp, tag, qtag->proto);
1427 if_printf(ife, "unable to prepend 802.1Q header");
1435 * Allocate an address from the FreeBSD Foundation OUI. This uses a
1436 * cryptographic hash function on the containing jail's name, UUID and the
1437 * interface name to attempt to provide a unique but stable address.
1438 * Pseudo-interfaces which require a MAC address should use this function to
1439 * allocate non-locally-administered addresses.
1442 ether_gen_addr(struct ifnet *ifp, struct ether_addr *hwaddr)
1446 char uuid[HOSTUUIDLEN + 1];
1449 char digest[SHA1_RESULTLEN];
1450 char jailname[MAXHOSTNAMELEN];
1452 getcredhostuuid(curthread->td_ucred, uuid, sizeof(uuid));
1453 if (strncmp(uuid, DEFAULT_HOSTUUID, sizeof(uuid)) == 0) {
1454 /* Fall back to a random mac address. */
1458 /* If each (vnet) jail would also have a unique hostuuid this would not
1460 getjailname(curthread->td_ucred, jailname, sizeof(jailname));
1461 sz = asprintf(&buf, M_TEMP, "%s-%s-%s", uuid, if_name(ifp),
1464 /* Fall back to a random mac address. */
1469 SHA1Update(&ctx, buf, sz);
1470 SHA1Final(digest, &ctx);
1473 addr = ((digest[0] << 16) | (digest[1] << 8) | digest[2]) &
1474 OUI_FREEBSD_GENERATED_MASK;
1475 addr = OUI_FREEBSD(addr);
1476 for (i = 0; i < ETHER_ADDR_LEN; ++i) {
1477 hwaddr->octet[i] = addr >> ((ETHER_ADDR_LEN - i - 1) * 8) &
1483 arc4rand(hwaddr, sizeof(*hwaddr), 0);
1485 hwaddr->octet[0] &= 0xFE;
1486 /* Locally administered. */
1487 hwaddr->octet[0] |= 0x02;
1490 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1491 MODULE_VERSION(ether, 1);