2 * Copyright (c) 1982, 1989, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)if_ethersubr.c 8.1 (Berkeley) 6/10/93
34 #include "opt_inet6.h"
35 #include "opt_netgraph.h"
36 #include "opt_mbuf_profiling.h"
39 #include <sys/param.h>
40 #include <sys/systm.h>
42 #include <sys/eventhandler.h>
44 #include <sys/kernel.h>
46 #include <sys/malloc.h>
47 #include <sys/module.h>
51 #include <sys/random.h>
52 #include <sys/socket.h>
53 #include <sys/sockio.h>
54 #include <sys/sysctl.h>
57 #include <net/ieee_oui.h>
59 #include <net/if_var.h>
60 #include <net/if_arp.h>
61 #include <net/netisr.h>
62 #include <net/route.h>
63 #include <net/if_llc.h>
64 #include <net/if_dl.h>
65 #include <net/if_types.h>
67 #include <net/ethernet.h>
68 #include <net/if_bridgevar.h>
69 #include <net/if_vlan_var.h>
70 #include <net/if_llatbl.h>
72 #include <net/rss_config.h>
75 #include <netpfil/pf/pf_mtag.h>
77 #if defined(INET) || defined(INET6)
78 #include <netinet/in.h>
79 #include <netinet/in_var.h>
80 #include <netinet/if_ether.h>
81 #include <netinet/ip_carp.h>
82 #include <netinet/ip_var.h>
85 #include <netinet6/nd6.h>
87 #include <security/mac/mac_framework.h>
89 #include <crypto/sha1.h>
92 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2);
93 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN);
96 VNET_DEFINE(struct pfil_head, link_pfil_hook); /* Packet filter hooks */
98 /* netgraph node hooks for ng_ether(4) */
99 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
100 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
101 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
102 void (*ng_ether_attach_p)(struct ifnet *ifp);
103 void (*ng_ether_detach_p)(struct ifnet *ifp);
105 void (*vlan_input_p)(struct ifnet *, struct mbuf *);
107 /* if_bridge(4) support */
108 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
109 int (*bridge_output_p)(struct ifnet *, struct mbuf *,
110 struct sockaddr *, struct rtentry *);
111 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
113 /* if_lagg(4) support */
114 struct mbuf *(*lagg_input_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 **,
122 static void ether_reassign(struct ifnet *, struct vnet *, char *);
124 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);
206 ether_resolve_addr(struct ifnet *ifp, struct mbuf *m,
207 const struct sockaddr *dst, struct route *ro, u_char *phdr,
208 uint32_t *pflags, struct llentry **plle)
210 struct ether_header *eh;
211 uint32_t lleflags = 0;
213 #if defined(INET) || defined(INET6)
219 eh = (struct ether_header *)phdr;
221 switch (dst->sa_family) {
224 if ((m->m_flags & (M_BCAST | M_MCAST)) == 0)
225 error = arpresolve(ifp, 0, m, dst, phdr, &lleflags,
228 if (m->m_flags & M_BCAST)
229 memcpy(eh->ether_dhost, ifp->if_broadcastaddr,
232 const struct in_addr *a;
233 a = &(((const struct sockaddr_in *)dst)->sin_addr);
234 ETHER_MAP_IP_MULTICAST(a, eh->ether_dhost);
236 etype = htons(ETHERTYPE_IP);
237 memcpy(&eh->ether_type, &etype, sizeof(etype));
238 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
244 if ((m->m_flags & M_MCAST) == 0)
245 error = nd6_resolve(ifp, 0, m, dst, phdr, &lleflags,
248 const struct in6_addr *a6;
249 a6 = &(((const struct sockaddr_in6 *)dst)->sin6_addr);
250 ETHER_MAP_IPV6_MULTICAST(a6, eh->ether_dhost);
251 etype = htons(ETHERTYPE_IPV6);
252 memcpy(&eh->ether_type, &etype, sizeof(etype));
253 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
258 if_printf(ifp, "can't handle af%d\n", dst->sa_family);
261 return (EAFNOSUPPORT);
264 if (error == EHOSTDOWN) {
265 if (ro != NULL && (ro->ro_flags & RT_HAS_GW) != 0)
266 error = EHOSTUNREACH;
272 *pflags = RT_MAY_LOOP;
273 if (lleflags & LLE_IFADDR)
280 * Ethernet output routine.
281 * Encapsulate a packet of type family for the local net.
282 * Use trailer local net encapsulation if enough data in first
283 * packet leaves a multiple of 512 bytes of data in remainder.
286 ether_output(struct ifnet *ifp, struct mbuf *m,
287 const struct sockaddr *dst, struct route *ro)
290 char linkhdr[ETHER_HDR_LEN], *phdr;
291 struct ether_header *eh;
294 int hlen; /* link layer header length */
296 struct llentry *lle = NULL;
297 struct rtentry *rt0 = NULL;
303 /* XXX BPF uses ro_prepend */
304 if (ro->ro_prepend != NULL) {
305 phdr = ro->ro_prepend;
307 } else if (!(m->m_flags & (M_BCAST | M_MCAST))) {
308 if ((ro->ro_flags & RT_LLE_CACHE) != 0) {
311 (lle->la_flags & LLE_VALID) == 0) {
313 lle = NULL; /* redundant */
317 /* if we lookup, keep cache */
321 * Notify LLE code that
325 llentry_mark_used(lle);
328 phdr = lle->r_linkdata;
329 hlen = lle->r_hdrlen;
330 pflags = lle->r_flags;
337 error = mac_ifnet_check_transmit(ifp, m);
343 if (ifp->if_flags & IFF_MONITOR)
345 if (!((ifp->if_flags & IFF_UP) &&
346 (ifp->if_drv_flags & IFF_DRV_RUNNING)))
350 /* No prepend data supplied. Try to calculate ourselves. */
352 hlen = ETHER_HDR_LEN;
353 error = ether_resolve_addr(ifp, m, dst, ro, phdr, &pflags,
354 addref ? &lle : NULL);
355 if (addref && lle != NULL)
358 return (error == EWOULDBLOCK ? 0 : error);
361 if ((pflags & RT_L2_ME) != 0) {
362 update_mbuf_csumflags(m, m);
363 return (if_simloop(ifp, m, dst->sa_family, 0));
365 loop_copy = pflags & RT_MAY_LOOP;
368 * Add local net header. If no space in first mbuf,
371 * Note that we do prepend regardless of RT_HAS_HEADER flag.
372 * This is done because BPF code shifts m_data pointer
373 * to the end of ethernet header prior to calling if_output().
375 M_PREPEND(m, hlen, M_NOWAIT);
378 if ((pflags & RT_HAS_HEADER) == 0) {
379 eh = mtod(m, struct ether_header *);
380 memcpy(eh, phdr, hlen);
384 * If a simplex interface, and the packet is being sent to our
385 * Ethernet address or a broadcast address, loopback a copy.
386 * XXX To make a simplex device behave exactly like a duplex
387 * device, we should copy in the case of sending to our own
388 * ethernet address (thus letting the original actually appear
389 * on the wire). However, we don't do that here for security
390 * reasons and compatibility with the original behavior.
392 if ((m->m_flags & M_BCAST) && loop_copy && (ifp->if_flags & IFF_SIMPLEX) &&
393 ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
397 * Because if_simloop() modifies the packet, we need a
398 * writable copy through m_dup() instead of a readonly
399 * one as m_copy[m] would give us. The alternative would
400 * be to modify if_simloop() to handle the readonly mbuf,
401 * but performancewise it is mostly equivalent (trading
402 * extra data copying vs. extra locking).
404 * XXX This is a local workaround. A number of less
405 * often used kernel parts suffer from the same bug.
406 * See PR kern/105943 for a proposed general solution.
408 if ((n = m_dup(m, M_NOWAIT)) != NULL) {
409 update_mbuf_csumflags(m, n);
410 (void)if_simloop(ifp, n, dst->sa_family, hlen);
412 if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
416 * Bridges require special output handling.
418 if (ifp->if_bridge) {
419 BRIDGE_OUTPUT(ifp, m, error);
423 #if defined(INET) || defined(INET6)
425 (error = (*carp_output_p)(ifp, m, dst)))
429 /* Handle ng_ether(4) processing, if any */
430 if (ifp->if_l2com != NULL) {
431 KASSERT(ng_ether_output_p != NULL,
432 ("ng_ether_output_p is NULL"));
433 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
442 /* Continue with link-layer output */
443 return ether_output_frame(ifp, m);
447 ether_set_pcp(struct mbuf **mp, struct ifnet *ifp, uint8_t pcp)
449 struct ether_header *eh;
451 eh = mtod(*mp, struct ether_header *);
452 if (ntohs(eh->ether_type) == ETHERTYPE_VLAN ||
453 ether_8021q_frame(mp, ifp, ifp, 0, pcp))
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)
472 if (pcp != IFNET_PCP_NONE && !ether_set_pcp(&m, ifp, pcp))
475 if (PFIL_HOOKED(&V_link_pfil_hook)) {
476 error = pfil_run_hooks(&V_link_pfil_hook, &m, ifp,
486 * Queue message on interface, update output statistics if
487 * successful, and start output if interface not yet active.
489 return ((ifp->if_transmit)(ifp, m));
493 * Process a received Ethernet packet; the packet is in the
494 * mbuf chain m with the ethernet header at the front.
497 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
499 struct ether_header *eh;
502 if ((ifp->if_flags & IFF_UP) == 0) {
507 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
508 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
513 if (m->m_len < ETHER_HDR_LEN) {
514 /* XXX maybe should pullup? */
515 if_printf(ifp, "discard frame w/o leading ethernet "
516 "header (len %u pkt len %u)\n",
517 m->m_len, m->m_pkthdr.len);
518 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
522 eh = mtod(m, struct ether_header *);
523 etype = ntohs(eh->ether_type);
524 random_harvest_queue(m, sizeof(*m), 2, RANDOM_NET_ETHER);
526 CURVNET_SET_QUIET(ifp->if_vnet);
528 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
529 if (ETHER_IS_BROADCAST(eh->ether_dhost))
530 m->m_flags |= M_BCAST;
532 m->m_flags |= M_MCAST;
533 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
538 * Tag the mbuf with an appropriate MAC label before any other
539 * consumers can get to it.
541 mac_ifnet_create_mbuf(ifp, m);
545 * Give bpf a chance at the packet.
547 ETHER_BPF_MTAP(ifp, m);
550 * If the CRC is still on the packet, trim it off. We do this once
551 * and once only in case we are re-entered. Nothing else on the
552 * Ethernet receive path expects to see the FCS.
554 if (m->m_flags & M_HASFCS) {
555 m_adj(m, -ETHER_CRC_LEN);
556 m->m_flags &= ~M_HASFCS;
559 if (!(ifp->if_capenable & IFCAP_HWSTATS))
560 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
562 /* Allow monitor mode to claim this frame, after stats are updated. */
563 if (ifp->if_flags & IFF_MONITOR) {
569 /* Handle input from a lagg(4) port */
570 if (ifp->if_type == IFT_IEEE8023ADLAG) {
571 KASSERT(lagg_input_p != NULL,
572 ("%s: if_lagg not loaded!", __func__));
573 m = (*lagg_input_p)(ifp, m);
575 ifp = m->m_pkthdr.rcvif;
583 * If the hardware did not process an 802.1Q tag, do this now,
584 * to allow 802.1P priority frames to be passed to the main input
586 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels.
588 if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) {
589 struct ether_vlan_header *evl;
591 if (m->m_len < sizeof(*evl) &&
592 (m = m_pullup(m, sizeof(*evl))) == NULL) {
594 if_printf(ifp, "cannot pullup VLAN header\n");
596 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
601 evl = mtod(m, struct ether_vlan_header *);
602 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
603 m->m_flags |= M_VLANTAG;
605 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
606 ETHER_HDR_LEN - ETHER_TYPE_LEN);
607 m_adj(m, ETHER_VLAN_ENCAP_LEN);
608 eh = mtod(m, struct ether_header *);
611 M_SETFIB(m, ifp->if_fib);
613 /* Allow ng_ether(4) to claim this frame. */
614 if (ifp->if_l2com != NULL) {
615 KASSERT(ng_ether_input_p != NULL,
616 ("%s: ng_ether_input_p is NULL", __func__));
617 m->m_flags &= ~M_PROMISC;
618 (*ng_ether_input_p)(ifp, &m);
623 eh = mtod(m, struct ether_header *);
627 * Allow if_bridge(4) to claim this frame.
628 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
629 * and the frame should be delivered locally.
631 if (ifp->if_bridge != NULL) {
632 m->m_flags &= ~M_PROMISC;
633 BRIDGE_INPUT(ifp, m);
638 eh = mtod(m, struct ether_header *);
641 #if defined(INET) || defined(INET6)
643 * Clear M_PROMISC on frame so that carp(4) will see it when the
644 * mbuf flows up to Layer 3.
645 * FreeBSD's implementation of carp(4) uses the inprotosw
646 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
647 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
648 * is outside the scope of the M_PROMISC test below.
649 * TODO: Maintain a hash table of ethernet addresses other than
650 * ether_dhost which may be active on this ifp.
652 if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
653 m->m_flags &= ~M_PROMISC;
658 * If the frame received was not for our MAC address, set the
659 * M_PROMISC flag on the mbuf chain. The frame may need to
660 * be seen by the rest of the Ethernet input path in case of
661 * re-entry (e.g. bridge, vlan, netgraph) but should not be
662 * seen by upper protocol layers.
664 if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
665 bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
666 m->m_flags |= M_PROMISC;
674 * Ethernet input dispatch; by default, direct dispatch here regardless of
675 * global configuration. However, if RSS is enabled, hook up RSS affinity
676 * so that when deferred or hybrid dispatch is enabled, we can redistribute
679 * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or
680 * not it had already done work distribution via multi-queue. Then we could
681 * direct dispatch in the event load balancing was already complete and
682 * handle the case of interfaces with different capabilities better.
684 * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions
685 * at multiple layers?
687 * XXXRW: For now, enable all this only if RSS is compiled in, although it
688 * works fine without RSS. Need to characterise the performance overhead
689 * of the detour through the netisr code in the event the result is always
693 ether_nh_input(struct mbuf *m)
697 KASSERT(m->m_pkthdr.rcvif != NULL,
698 ("%s: NULL interface pointer", __func__));
699 ether_input_internal(m->m_pkthdr.rcvif, m);
702 static struct netisr_handler ether_nh = {
704 .nh_handler = ether_nh_input,
705 .nh_proto = NETISR_ETHER,
707 .nh_policy = NETISR_POLICY_CPU,
708 .nh_dispatch = NETISR_DISPATCH_DIRECT,
709 .nh_m2cpuid = rss_m2cpuid,
711 .nh_policy = NETISR_POLICY_SOURCE,
712 .nh_dispatch = NETISR_DISPATCH_DIRECT,
717 ether_init(__unused void *arg)
720 netisr_register(ðer_nh);
722 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
725 vnet_ether_init(__unused void *arg)
729 /* Initialize packet filter hooks. */
730 V_link_pfil_hook.ph_type = PFIL_TYPE_AF;
731 V_link_pfil_hook.ph_af = AF_LINK;
732 if ((i = pfil_head_register(&V_link_pfil_hook)) != 0)
733 printf("%s: WARNING: unable to register pfil link hook, "
734 "error %d\n", __func__, i);
736 netisr_register_vnet(ðer_nh);
739 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY,
740 vnet_ether_init, NULL);
744 vnet_ether_pfil_destroy(__unused void *arg)
748 if ((i = pfil_head_unregister(&V_link_pfil_hook)) != 0)
749 printf("%s: WARNING: unable to unregister pfil link hook, "
750 "error %d\n", __func__, i);
752 VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY,
753 vnet_ether_pfil_destroy, NULL);
756 vnet_ether_destroy(__unused void *arg)
759 netisr_unregister_vnet(ðer_nh);
761 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY,
762 vnet_ether_destroy, NULL);
768 ether_input(struct ifnet *ifp, struct mbuf *m)
774 * The drivers are allowed to pass in a chain of packets linked with
775 * m_nextpkt. We split them up into separate packets here and pass
776 * them up. This allows the drivers to amortize the receive lock.
783 * We will rely on rcvif being set properly in the deferred context,
784 * so assert it is correct here.
786 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p "
787 "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp));
788 CURVNET_SET_QUIET(ifp->if_vnet);
789 netisr_dispatch(NETISR_ETHER, m);
796 * Upper layer processing for a received Ethernet packet.
799 ether_demux(struct ifnet *ifp, struct mbuf *m)
801 struct ether_header *eh;
805 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
807 /* Do not grab PROMISC frames in case we are re-entered. */
808 if (PFIL_HOOKED(&V_link_pfil_hook) && !(m->m_flags & M_PROMISC)) {
809 i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_IN, 0,
812 if (i != 0 || m == NULL)
816 eh = mtod(m, struct ether_header *);
817 ether_type = ntohs(eh->ether_type);
820 * If this frame has a VLAN tag other than 0, call vlan_input()
821 * if its module is loaded. Otherwise, drop.
823 if ((m->m_flags & M_VLANTAG) &&
824 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
825 if (ifp->if_vlantrunk == NULL) {
826 if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
830 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
832 /* Clear before possibly re-entering ether_input(). */
833 m->m_flags &= ~M_PROMISC;
834 (*vlan_input_p)(ifp, m);
839 * Pass promiscuously received frames to the upper layer if the user
840 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
842 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
848 * Reset layer specific mbuf flags to avoid confusing upper layers.
849 * Strip off Ethernet header.
851 m->m_flags &= ~M_VLANTAG;
853 m_adj(m, ETHER_HDR_LEN);
856 * Dispatch frame to upper layer.
858 switch (ether_type) {
865 if (ifp->if_flags & IFF_NOARP) {
866 /* Discard packet if ARP is disabled on interface */
881 netisr_dispatch(isr, m);
886 * Packet is to be discarded. If netgraph is present,
887 * hand the packet to it for last chance processing;
888 * otherwise dispose of it.
890 if (ifp->if_l2com != NULL) {
891 KASSERT(ng_ether_input_orphan_p != NULL,
892 ("ng_ether_input_orphan_p is NULL"));
894 * Put back the ethernet header so netgraph has a
895 * consistent view of inbound packets.
897 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
898 (*ng_ether_input_orphan_p)(ifp, m);
905 * Convert Ethernet address to printable (loggable) representation.
906 * This routine is for compatibility; it's better to just use
908 * printf("%6D", <pointer to address>, ":");
910 * since there's no static buffer involved.
913 ether_sprintf(const u_char *ap)
915 static char etherbuf[18];
916 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
921 * Perform common duties while attaching to interface list
924 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
928 struct sockaddr_dl *sdl;
930 ifp->if_addrlen = ETHER_ADDR_LEN;
931 ifp->if_hdrlen = ETHER_HDR_LEN;
933 ifp->if_mtu = ETHERMTU;
934 ifp->if_output = ether_output;
935 ifp->if_input = ether_input;
936 ifp->if_resolvemulti = ether_resolvemulti;
937 ifp->if_requestencap = ether_requestencap;
939 ifp->if_reassign = ether_reassign;
941 if (ifp->if_baudrate == 0)
942 ifp->if_baudrate = IF_Mbps(10); /* just a default */
943 ifp->if_broadcastaddr = etherbroadcastaddr;
946 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
947 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
948 sdl->sdl_type = IFT_ETHER;
949 sdl->sdl_alen = ifp->if_addrlen;
950 bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
952 if (ifp->if_hw_addr != NULL)
953 bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen);
955 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
956 if (ng_ether_attach_p != NULL)
957 (*ng_ether_attach_p)(ifp);
959 /* Announce Ethernet MAC address if non-zero. */
960 for (i = 0; i < ifp->if_addrlen; i++)
963 if (i != ifp->if_addrlen)
964 if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
966 uuid_ether_add(LLADDR(sdl));
968 /* Add necessary bits are setup; announce it now. */
969 EVENTHANDLER_INVOKE(ether_ifattach_event, ifp);
970 if (IS_DEFAULT_VNET(curvnet))
971 devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL);
975 * Perform common duties while detaching an Ethernet interface
978 ether_ifdetach(struct ifnet *ifp)
980 struct sockaddr_dl *sdl;
982 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr);
983 uuid_ether_del(LLADDR(sdl));
985 if (ifp->if_l2com != NULL) {
986 KASSERT(ng_ether_detach_p != NULL,
987 ("ng_ether_detach_p is NULL"));
988 (*ng_ether_detach_p)(ifp);
997 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused)
1000 if (ifp->if_l2com != NULL) {
1001 KASSERT(ng_ether_detach_p != NULL,
1002 ("ng_ether_detach_p is NULL"));
1003 (*ng_ether_detach_p)(ifp);
1006 if (ng_ether_attach_p != NULL) {
1007 CURVNET_SET_QUIET(new_vnet);
1008 (*ng_ether_attach_p)(ifp);
1014 SYSCTL_DECL(_net_link);
1015 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
1019 * This is for reference. We have a table-driven version
1020 * of the little-endian crc32 generator, which is faster
1021 * than the double-loop.
1024 ether_crc32_le(const uint8_t *buf, size_t len)
1031 crc = 0xffffffff; /* initial value */
1033 for (i = 0; i < len; i++) {
1034 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1035 carry = (crc ^ data) & 1;
1038 crc = (crc ^ ETHER_CRC_POLY_LE);
1046 ether_crc32_le(const uint8_t *buf, size_t len)
1048 static const uint32_t crctab[] = {
1049 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1050 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1051 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1052 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1057 crc = 0xffffffff; /* initial value */
1059 for (i = 0; i < len; i++) {
1061 crc = (crc >> 4) ^ crctab[crc & 0xf];
1062 crc = (crc >> 4) ^ crctab[crc & 0xf];
1070 ether_crc32_be(const uint8_t *buf, size_t len)
1073 uint32_t crc, carry;
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 & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1084 crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1092 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1094 struct ifaddr *ifa = (struct ifaddr *) data;
1095 struct ifreq *ifr = (struct ifreq *) data;
1100 ifp->if_flags |= IFF_UP;
1102 switch (ifa->ifa_addr->sa_family) {
1105 ifp->if_init(ifp->if_softc); /* before arpwhohas */
1106 arp_ifinit(ifp, ifa);
1110 ifp->if_init(ifp->if_softc);
1116 bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0],
1122 * Set the interface MTU.
1124 if (ifr->ifr_mtu > ETHERMTU) {
1127 ifp->if_mtu = ifr->ifr_mtu;
1132 error = priv_check(curthread, PRIV_NET_SETLANPCP);
1135 if (ifr->ifr_lan_pcp > 7 &&
1136 ifr->ifr_lan_pcp != IFNET_PCP_NONE) {
1139 ifp->if_pcp = ifr->ifr_lan_pcp;
1140 /* broadcast event about PCP change */
1141 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP);
1146 ifr->ifr_lan_pcp = ifp->if_pcp;
1150 error = EINVAL; /* XXX netbsd has ENOTTY??? */
1157 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1158 struct sockaddr *sa)
1160 struct sockaddr_dl *sdl;
1162 struct sockaddr_in *sin;
1165 struct sockaddr_in6 *sin6;
1169 switch(sa->sa_family) {
1172 * No mapping needed. Just check that it's a valid MC address.
1174 sdl = (struct sockaddr_dl *)sa;
1175 e_addr = LLADDR(sdl);
1176 if (!ETHER_IS_MULTICAST(e_addr))
1177 return EADDRNOTAVAIL;
1183 sin = (struct sockaddr_in *)sa;
1184 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1185 return EADDRNOTAVAIL;
1186 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1187 sdl->sdl_alen = ETHER_ADDR_LEN;
1188 e_addr = LLADDR(sdl);
1189 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1190 *llsa = (struct sockaddr *)sdl;
1195 sin6 = (struct sockaddr_in6 *)sa;
1196 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1198 * An IP6 address of 0 means listen to all
1199 * of the Ethernet multicast address used for IP6.
1200 * (This is used for multicast routers.)
1202 ifp->if_flags |= IFF_ALLMULTI;
1206 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1207 return EADDRNOTAVAIL;
1208 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1209 sdl->sdl_alen = ETHER_ADDR_LEN;
1210 e_addr = LLADDR(sdl);
1211 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1212 *llsa = (struct sockaddr *)sdl;
1218 * Well, the text isn't quite right, but it's the name
1221 return EAFNOSUPPORT;
1225 static moduledata_t ether_mod = {
1230 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1232 struct ether_vlan_header vlan;
1235 KASSERT((m->m_flags & M_VLANTAG) != 0,
1236 ("%s: vlan information not present", __func__));
1237 KASSERT(m->m_len >= sizeof(struct ether_header),
1238 ("%s: mbuf not large enough for header", __func__));
1239 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1240 vlan.evl_proto = vlan.evl_encap_proto;
1241 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1242 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1243 m->m_len -= sizeof(struct ether_header);
1244 m->m_data += sizeof(struct ether_header);
1246 * If a data link has been supplied by the caller, then we will need to
1247 * re-create a stack allocated mbuf chain with the following structure:
1249 * (1) mbuf #1 will contain the supplied data link
1250 * (2) mbuf #2 will contain the vlan header
1251 * (3) mbuf #3 will contain the original mbuf's packet data
1253 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1257 mv.m_data = (caddr_t)&vlan;
1258 mv.m_len = sizeof(vlan);
1264 bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1265 m->m_len += sizeof(struct ether_header);
1266 m->m_data -= sizeof(struct ether_header);
1270 ether_vlanencap(struct mbuf *m, uint16_t tag)
1272 struct ether_vlan_header *evl;
1274 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT);
1277 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1279 if (m->m_len < sizeof(*evl)) {
1280 m = m_pullup(m, sizeof(*evl));
1286 * Transform the Ethernet header into an Ethernet header
1287 * with 802.1Q encapsulation.
1289 evl = mtod(m, struct ether_vlan_header *);
1290 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1291 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1292 evl->evl_encap_proto = htons(ETHERTYPE_VLAN);
1293 evl->evl_tag = htons(tag);
1297 static SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW, 0,
1298 "IEEE 802.1Q VLAN");
1299 static SYSCTL_NODE(_net_link_vlan, PF_LINK, link, CTLFLAG_RW, 0,
1302 static VNET_DEFINE(int, soft_pad);
1303 #define V_soft_pad VNET(soft_pad)
1304 SYSCTL_INT(_net_link_vlan, OID_AUTO, soft_pad, CTLFLAG_RW | CTLFLAG_VNET,
1305 &VNET_NAME(soft_pad), 0,
1306 "pad short frames before tagging");
1309 * For now, make preserving PCP via an mbuf tag optional, as it increases
1310 * per-packet memory allocations and frees. In the future, it would be
1311 * preferable to reuse ether_vtag for this, or similar.
1313 int vlan_mtag_pcp = 0;
1314 SYSCTL_INT(_net_link_vlan, OID_AUTO, mtag_pcp, CTLFLAG_RW,
1316 "Retain VLAN PCP information as packets are passed up the stack");
1319 ether_8021q_frame(struct mbuf **mp, struct ifnet *ife, struct ifnet *p,
1320 uint16_t vid, uint8_t pcp)
1325 static const char pad[8]; /* just zeros */
1328 * Pad the frame to the minimum size allowed if told to.
1329 * This option is in accord with IEEE Std 802.1Q, 2003 Ed.,
1330 * paragraph C.4.4.3.b. It can help to work around buggy
1331 * bridges that violate paragraph C.4.4.3.a from the same
1332 * document, i.e., fail to pad short frames after untagging.
1333 * E.g., a tagged frame 66 bytes long (incl. FCS) is OK, but
1334 * untagging it will produce a 62-byte frame, which is a runt
1335 * and requires padding. There are VLAN-enabled network
1336 * devices that just discard such runts instead or mishandle
1339 if (V_soft_pad && p->if_type == IFT_ETHER) {
1340 for (n = ETHERMIN + ETHER_HDR_LEN - (*mp)->m_pkthdr.len;
1341 n > 0; n -= sizeof(pad)) {
1342 if (!m_append(*mp, min(n, sizeof(pad)), pad))
1348 if_printf(ife, "cannot pad short frame");
1354 * If underlying interface can do VLAN tag insertion itself,
1355 * just pass the packet along. However, we need some way to
1356 * tell the interface where the packet came from so that it
1357 * knows how to find the VLAN tag to use, so we attach a
1358 * packet tag that holds it.
1360 if (vlan_mtag_pcp && (mtag = m_tag_locate(*mp, MTAG_8021Q,
1361 MTAG_8021Q_PCP_OUT, NULL)) != NULL)
1362 tag = EVL_MAKETAG(vid, *(uint8_t *)(mtag + 1), 0);
1364 tag = EVL_MAKETAG(vid, pcp, 0);
1365 if (p->if_capenable & IFCAP_VLAN_HWTAGGING) {
1366 (*mp)->m_pkthdr.ether_vtag = tag;
1367 (*mp)->m_flags |= M_VLANTAG;
1369 *mp = ether_vlanencap(*mp, tag);
1371 if_printf(ife, "unable to prepend 802.1Q header");
1379 * Allocate an address from the FreeBSD Foundation OUI. This uses a
1380 * cryptographic hash function on the containing jail's UUID and the interface
1381 * name to attempt to provide a unique but stable address. Pseudo-interfaces
1382 * which require a MAC address should use this function to allocate
1383 * non-locally-administered addresses.
1386 ether_gen_addr(struct ifnet *ifp, struct ether_addr *hwaddr)
1388 #define ETHER_GEN_ADDR_BUFSIZ HOSTUUIDLEN + IFNAMSIZ + 2
1390 char buf[ETHER_GEN_ADDR_BUFSIZ];
1391 char uuid[HOSTUUIDLEN + 1];
1394 char digest[SHA1_RESULTLEN];
1396 getcredhostuuid(curthread->td_ucred, uuid, sizeof(uuid));
1397 sz = snprintf(buf, ETHER_GEN_ADDR_BUFSIZ, "%s-%s", uuid, ifp->if_xname);
1399 SHA1Update(&ctx, buf, sz);
1400 SHA1Final(digest, &ctx);
1402 addr = ((digest[0] << 16) | (digest[1] << 8) | digest[2]) &
1403 OUI_FREEBSD_GENERATED_MASK;
1404 addr = OUI_FREEBSD(addr);
1405 for (i = 0; i < ETHER_ADDR_LEN; ++i) {
1406 hwaddr->octet[i] = addr >> ((ETHER_ADDR_LEN - i - 1) * 8) &
1411 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1412 MODULE_VERSION(ether, 1);