2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1989, 1993
5 * The Regents of the University of California. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * @(#)if_ethersubr.c 8.1 (Berkeley) 6/10/93
36 #include "opt_inet6.h"
37 #include "opt_netgraph.h"
38 #include "opt_mbuf_profiling.h"
41 #include <sys/param.h>
42 #include <sys/systm.h>
44 #include <sys/eventhandler.h>
45 #include <sys/kernel.h>
47 #include <sys/malloc.h>
48 #include <sys/module.h>
51 #include <sys/random.h>
52 #include <sys/socket.h>
53 #include <sys/sockio.h>
54 #include <sys/sysctl.h>
58 #include <net/if_var.h>
59 #include <net/if_arp.h>
60 #include <net/netisr.h>
61 #include <net/route.h>
62 #include <net/if_llc.h>
63 #include <net/if_dl.h>
64 #include <net/if_types.h>
66 #include <net/ethernet.h>
67 #include <net/if_bridgevar.h>
68 #include <net/if_vlan_var.h>
69 #include <net/if_llatbl.h>
71 #include <net/rss_config.h>
74 #include <netpfil/pf/pf_mtag.h>
76 #if defined(INET) || defined(INET6)
77 #include <netinet/in.h>
78 #include <netinet/in_var.h>
79 #include <netinet/if_ether.h>
80 #include <netinet/ip_carp.h>
81 #include <netinet/ip_var.h>
84 #include <netinet6/nd6.h>
86 #include <security/mac/mac_framework.h>
89 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2);
90 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN);
93 VNET_DEFINE(struct pfil_head, link_pfil_hook); /* Packet filter hooks */
95 /* netgraph node hooks for ng_ether(4) */
96 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
97 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
98 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
99 void (*ng_ether_attach_p)(struct ifnet *ifp);
100 void (*ng_ether_detach_p)(struct ifnet *ifp);
102 void (*vlan_input_p)(struct ifnet *, struct mbuf *);
104 /* if_bridge(4) support */
105 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
107 /* if_lagg(4) support */
108 struct mbuf *(*lagg_input_p)(struct ifnet *, struct mbuf *);
110 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
111 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
113 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
116 static void ether_reassign(struct ifnet *, struct vnet *, char *);
118 static int ether_requestencap(struct ifnet *, struct if_encap_req *);
121 #define senderr(e) do { error = (e); goto bad;} while (0)
124 update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst)
128 if (src->m_pkthdr.csum_flags & CSUM_IP)
129 csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
130 if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
131 csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
132 if (src->m_pkthdr.csum_flags & CSUM_SCTP)
133 csum_flags |= CSUM_SCTP_VALID;
134 dst->m_pkthdr.csum_flags |= csum_flags;
135 if (csum_flags & CSUM_DATA_VALID)
136 dst->m_pkthdr.csum_data = 0xffff;
140 * Handle link-layer encapsulation requests.
143 ether_requestencap(struct ifnet *ifp, struct if_encap_req *req)
145 struct ether_header *eh;
148 const u_char *lladdr;
150 if (req->rtype != IFENCAP_LL)
153 if (req->bufsize < ETHER_HDR_LEN)
156 eh = (struct ether_header *)req->buf;
157 lladdr = req->lladdr;
160 switch (req->family) {
162 etype = htons(ETHERTYPE_IP);
165 etype = htons(ETHERTYPE_IPV6);
168 ah = (struct arphdr *)req->hdata;
169 ah->ar_hrd = htons(ARPHRD_ETHER);
171 switch(ntohs(ah->ar_op)) {
172 case ARPOP_REVREQUEST:
174 etype = htons(ETHERTYPE_REVARP);
179 etype = htons(ETHERTYPE_ARP);
183 if (req->flags & IFENCAP_FLAG_BROADCAST)
184 lladdr = ifp->if_broadcastaddr;
187 return (EAFNOSUPPORT);
190 memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
191 memcpy(eh->ether_dhost, lladdr, ETHER_ADDR_LEN);
192 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
193 req->bufsize = sizeof(struct ether_header);
200 ether_resolve_addr(struct ifnet *ifp, struct mbuf *m,
201 const struct sockaddr *dst, struct route *ro, u_char *phdr,
202 uint32_t *pflags, struct llentry **plle)
204 struct ether_header *eh;
205 uint32_t lleflags = 0;
207 #if defined(INET) || defined(INET6)
213 eh = (struct ether_header *)phdr;
215 switch (dst->sa_family) {
218 if ((m->m_flags & (M_BCAST | M_MCAST)) == 0)
219 error = arpresolve(ifp, 0, m, dst, phdr, &lleflags,
222 if (m->m_flags & M_BCAST)
223 memcpy(eh->ether_dhost, ifp->if_broadcastaddr,
226 const struct in_addr *a;
227 a = &(((const struct sockaddr_in *)dst)->sin_addr);
228 ETHER_MAP_IP_MULTICAST(a, eh->ether_dhost);
230 etype = htons(ETHERTYPE_IP);
231 memcpy(&eh->ether_type, &etype, sizeof(etype));
232 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
238 if ((m->m_flags & M_MCAST) == 0)
239 error = nd6_resolve(ifp, 0, m, dst, phdr, &lleflags,
242 const struct in6_addr *a6;
243 a6 = &(((const struct sockaddr_in6 *)dst)->sin6_addr);
244 ETHER_MAP_IPV6_MULTICAST(a6, eh->ether_dhost);
245 etype = htons(ETHERTYPE_IPV6);
246 memcpy(&eh->ether_type, &etype, sizeof(etype));
247 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
252 if_printf(ifp, "can't handle af%d\n", dst->sa_family);
255 return (EAFNOSUPPORT);
258 if (error == EHOSTDOWN) {
259 if (ro != NULL && (ro->ro_flags & RT_HAS_GW) != 0)
260 error = EHOSTUNREACH;
266 *pflags = RT_MAY_LOOP;
267 if (lleflags & LLE_IFADDR)
274 * Ethernet output routine.
275 * Encapsulate a packet of type family for the local net.
276 * Use trailer local net encapsulation if enough data in first
277 * packet leaves a multiple of 512 bytes of data in remainder.
280 ether_output(struct ifnet *ifp, struct mbuf *m,
281 const struct sockaddr *dst, struct route *ro)
284 char linkhdr[ETHER_HDR_LEN], *phdr;
285 struct ether_header *eh;
288 int hlen; /* link layer header length */
290 struct llentry *lle = NULL;
296 /* XXX BPF uses ro_prepend */
297 if (ro->ro_prepend != NULL) {
298 phdr = ro->ro_prepend;
300 } else if (!(m->m_flags & (M_BCAST | M_MCAST))) {
301 if ((ro->ro_flags & RT_LLE_CACHE) != 0) {
304 (lle->la_flags & LLE_VALID) == 0) {
306 lle = NULL; /* redundant */
310 /* if we lookup, keep cache */
314 * Notify LLE code that
318 llentry_mark_used(lle);
321 phdr = lle->r_linkdata;
322 hlen = lle->r_hdrlen;
323 pflags = lle->r_flags;
329 error = mac_ifnet_check_transmit(ifp, m);
335 if (ifp->if_flags & IFF_MONITOR)
337 if (!((ifp->if_flags & IFF_UP) &&
338 (ifp->if_drv_flags & IFF_DRV_RUNNING)))
342 /* No prepend data supplied. Try to calculate ourselves. */
344 hlen = ETHER_HDR_LEN;
345 error = ether_resolve_addr(ifp, m, dst, ro, phdr, &pflags,
346 addref ? &lle : NULL);
347 if (addref && lle != NULL)
350 return (error == EWOULDBLOCK ? 0 : error);
353 if ((pflags & RT_L2_ME) != 0) {
354 update_mbuf_csumflags(m, m);
355 return (if_simloop(ifp, m, dst->sa_family, 0));
357 loop_copy = pflags & RT_MAY_LOOP;
360 * Add local net header. If no space in first mbuf,
363 * Note that we do prepend regardless of RT_HAS_HEADER flag.
364 * This is done because BPF code shifts m_data pointer
365 * to the end of ethernet header prior to calling if_output().
367 M_PREPEND(m, hlen, M_NOWAIT);
370 if ((pflags & RT_HAS_HEADER) == 0) {
371 eh = mtod(m, struct ether_header *);
372 memcpy(eh, phdr, hlen);
376 * If a simplex interface, and the packet is being sent to our
377 * Ethernet address or a broadcast address, loopback a copy.
378 * XXX To make a simplex device behave exactly like a duplex
379 * device, we should copy in the case of sending to our own
380 * ethernet address (thus letting the original actually appear
381 * on the wire). However, we don't do that here for security
382 * reasons and compatibility with the original behavior.
384 if ((m->m_flags & M_BCAST) && loop_copy && (ifp->if_flags & IFF_SIMPLEX) &&
385 ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
389 * Because if_simloop() modifies the packet, we need a
390 * writable copy through m_dup() instead of a readonly
391 * one as m_copy[m] would give us. The alternative would
392 * be to modify if_simloop() to handle the readonly mbuf,
393 * but performancewise it is mostly equivalent (trading
394 * extra data copying vs. extra locking).
396 * XXX This is a local workaround. A number of less
397 * often used kernel parts suffer from the same bug.
398 * See PR kern/105943 for a proposed general solution.
400 if ((n = m_dup(m, M_NOWAIT)) != NULL) {
401 update_mbuf_csumflags(m, n);
402 (void)if_simloop(ifp, n, dst->sa_family, hlen);
404 if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
408 * Bridges require special output handling.
410 if (ifp->if_bridge) {
411 BRIDGE_OUTPUT(ifp, m, error);
415 #if defined(INET) || defined(INET6)
417 (error = (*carp_output_p)(ifp, m, dst)))
421 /* Handle ng_ether(4) processing, if any */
422 if (ifp->if_l2com != NULL) {
423 KASSERT(ng_ether_output_p != NULL,
424 ("ng_ether_output_p is NULL"));
425 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
434 /* Continue with link-layer output */
435 return ether_output_frame(ifp, m);
439 ether_set_pcp(struct mbuf **mp, struct ifnet *ifp, uint8_t pcp)
441 struct ether_header *eh;
443 eh = mtod(*mp, struct ether_header *);
444 if (ntohs(eh->ether_type) == ETHERTYPE_VLAN ||
445 ether_8021q_frame(mp, ifp, ifp, 0, pcp))
447 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
452 * Ethernet link layer output routine to send a raw frame to the device.
454 * This assumes that the 14 byte Ethernet header is present and contiguous
455 * in the first mbuf (if BRIDGE'ing).
458 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
464 if (pcp != IFNET_PCP_NONE && ifp->if_type != IFT_L2VLAN &&
465 !ether_set_pcp(&m, ifp, pcp))
468 if (PFIL_HOOKED(&V_link_pfil_hook)) {
469 error = pfil_run_hooks(&V_link_pfil_hook, &m, ifp,
479 #if defined(INET6) && defined(INET)
480 /* draft-ietf-6man-ipv6only-flag */
481 /* Catch ETHERTYPE_IP, and ETHERTYPE_ARP if we are v6-only. */
482 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY) != 0) {
483 struct ether_header *eh;
485 eh = mtod(m, struct ether_header *);
486 switch (ntohs(eh->ether_type)) {
490 return (EAFNOSUPPORT);
499 * Queue message on interface, update output statistics if
500 * successful, and start output if interface not yet active.
502 return ((ifp->if_transmit)(ifp, m));
506 * Process a received Ethernet packet; the packet is in the
507 * mbuf chain m with the ethernet header at the front.
510 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
512 struct ether_header *eh;
515 if ((ifp->if_flags & IFF_UP) == 0) {
520 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
521 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
526 if (m->m_len < ETHER_HDR_LEN) {
527 /* XXX maybe should pullup? */
528 if_printf(ifp, "discard frame w/o leading ethernet "
529 "header (len %u pkt len %u)\n",
530 m->m_len, m->m_pkthdr.len);
531 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
535 eh = mtod(m, struct ether_header *);
536 etype = ntohs(eh->ether_type);
537 random_harvest_queue_ether(m, sizeof(*m));
539 CURVNET_SET_QUIET(ifp->if_vnet);
541 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
542 if (ETHER_IS_BROADCAST(eh->ether_dhost))
543 m->m_flags |= M_BCAST;
545 m->m_flags |= M_MCAST;
546 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
551 * Tag the mbuf with an appropriate MAC label before any other
552 * consumers can get to it.
554 mac_ifnet_create_mbuf(ifp, m);
558 * Give bpf a chance at the packet.
560 ETHER_BPF_MTAP(ifp, m);
563 * If the CRC is still on the packet, trim it off. We do this once
564 * and once only in case we are re-entered. Nothing else on the
565 * Ethernet receive path expects to see the FCS.
567 if (m->m_flags & M_HASFCS) {
568 m_adj(m, -ETHER_CRC_LEN);
569 m->m_flags &= ~M_HASFCS;
572 if (!(ifp->if_capenable & IFCAP_HWSTATS))
573 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
575 /* Allow monitor mode to claim this frame, after stats are updated. */
576 if (ifp->if_flags & IFF_MONITOR) {
582 /* Handle input from a lagg(4) port */
583 if (ifp->if_type == IFT_IEEE8023ADLAG) {
584 KASSERT(lagg_input_p != NULL,
585 ("%s: if_lagg not loaded!", __func__));
586 m = (*lagg_input_p)(ifp, m);
588 ifp = m->m_pkthdr.rcvif;
596 * If the hardware did not process an 802.1Q tag, do this now,
597 * to allow 802.1P priority frames to be passed to the main input
599 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels.
601 if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) {
602 struct ether_vlan_header *evl;
604 if (m->m_len < sizeof(*evl) &&
605 (m = m_pullup(m, sizeof(*evl))) == NULL) {
607 if_printf(ifp, "cannot pullup VLAN header\n");
609 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
614 evl = mtod(m, struct ether_vlan_header *);
615 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
616 m->m_flags |= M_VLANTAG;
618 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
619 ETHER_HDR_LEN - ETHER_TYPE_LEN);
620 m_adj(m, ETHER_VLAN_ENCAP_LEN);
621 eh = mtod(m, struct ether_header *);
624 M_SETFIB(m, ifp->if_fib);
626 /* Allow ng_ether(4) to claim this frame. */
627 if (ifp->if_l2com != NULL) {
628 KASSERT(ng_ether_input_p != NULL,
629 ("%s: ng_ether_input_p is NULL", __func__));
630 m->m_flags &= ~M_PROMISC;
631 (*ng_ether_input_p)(ifp, &m);
636 eh = mtod(m, struct ether_header *);
640 * Allow if_bridge(4) to claim this frame.
641 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
642 * and the frame should be delivered locally.
644 if (ifp->if_bridge != NULL) {
645 m->m_flags &= ~M_PROMISC;
646 BRIDGE_INPUT(ifp, m);
651 eh = mtod(m, struct ether_header *);
654 #if defined(INET) || defined(INET6)
656 * Clear M_PROMISC on frame so that carp(4) will see it when the
657 * mbuf flows up to Layer 3.
658 * FreeBSD's implementation of carp(4) uses the inprotosw
659 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
660 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
661 * is outside the scope of the M_PROMISC test below.
662 * TODO: Maintain a hash table of ethernet addresses other than
663 * ether_dhost which may be active on this ifp.
665 if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
666 m->m_flags &= ~M_PROMISC;
671 * If the frame received was not for our MAC address, set the
672 * M_PROMISC flag on the mbuf chain. The frame may need to
673 * be seen by the rest of the Ethernet input path in case of
674 * re-entry (e.g. bridge, vlan, netgraph) but should not be
675 * seen by upper protocol layers.
677 if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
678 bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
679 m->m_flags |= M_PROMISC;
687 * Ethernet input dispatch; by default, direct dispatch here regardless of
688 * global configuration. However, if RSS is enabled, hook up RSS affinity
689 * so that when deferred or hybrid dispatch is enabled, we can redistribute
692 * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or
693 * not it had already done work distribution via multi-queue. Then we could
694 * direct dispatch in the event load balancing was already complete and
695 * handle the case of interfaces with different capabilities better.
697 * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions
698 * at multiple layers?
700 * XXXRW: For now, enable all this only if RSS is compiled in, although it
701 * works fine without RSS. Need to characterise the performance overhead
702 * of the detour through the netisr code in the event the result is always
706 ether_nh_input(struct mbuf *m)
710 KASSERT(m->m_pkthdr.rcvif != NULL,
711 ("%s: NULL interface pointer", __func__));
712 ether_input_internal(m->m_pkthdr.rcvif, m);
715 static struct netisr_handler ether_nh = {
717 .nh_handler = ether_nh_input,
718 .nh_proto = NETISR_ETHER,
720 .nh_policy = NETISR_POLICY_CPU,
721 .nh_dispatch = NETISR_DISPATCH_DIRECT,
722 .nh_m2cpuid = rss_m2cpuid,
724 .nh_policy = NETISR_POLICY_SOURCE,
725 .nh_dispatch = NETISR_DISPATCH_DIRECT,
730 ether_init(__unused void *arg)
733 netisr_register(ðer_nh);
735 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
738 vnet_ether_init(__unused void *arg)
742 /* Initialize packet filter hooks. */
743 V_link_pfil_hook.ph_type = PFIL_TYPE_AF;
744 V_link_pfil_hook.ph_af = AF_LINK;
745 if ((i = pfil_head_register(&V_link_pfil_hook)) != 0)
746 printf("%s: WARNING: unable to register pfil link hook, "
747 "error %d\n", __func__, i);
749 netisr_register_vnet(ðer_nh);
752 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY,
753 vnet_ether_init, NULL);
757 vnet_ether_pfil_destroy(__unused void *arg)
761 if ((i = pfil_head_unregister(&V_link_pfil_hook)) != 0)
762 printf("%s: WARNING: unable to unregister pfil link hook, "
763 "error %d\n", __func__, i);
765 VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY,
766 vnet_ether_pfil_destroy, NULL);
769 vnet_ether_destroy(__unused void *arg)
772 netisr_unregister_vnet(ðer_nh);
774 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY,
775 vnet_ether_destroy, NULL);
781 ether_input(struct ifnet *ifp, struct mbuf *m)
787 * The drivers are allowed to pass in a chain of packets linked with
788 * m_nextpkt. We split them up into separate packets here and pass
789 * them up. This allows the drivers to amortize the receive lock.
796 * We will rely on rcvif being set properly in the deferred context,
797 * so assert it is correct here.
799 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p "
800 "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp));
801 CURVNET_SET_QUIET(ifp->if_vnet);
802 netisr_dispatch(NETISR_ETHER, m);
809 * Upper layer processing for a received Ethernet packet.
812 ether_demux(struct ifnet *ifp, struct mbuf *m)
814 struct ether_header *eh;
818 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
820 /* Do not grab PROMISC frames in case we are re-entered. */
821 if (PFIL_HOOKED(&V_link_pfil_hook) && !(m->m_flags & M_PROMISC)) {
822 i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_IN, 0,
825 if (i != 0 || m == NULL)
829 eh = mtod(m, struct ether_header *);
830 ether_type = ntohs(eh->ether_type);
833 * If this frame has a VLAN tag other than 0, call vlan_input()
834 * if its module is loaded. Otherwise, drop.
836 if ((m->m_flags & M_VLANTAG) &&
837 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
838 if (ifp->if_vlantrunk == NULL) {
839 if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
843 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
845 /* Clear before possibly re-entering ether_input(). */
846 m->m_flags &= ~M_PROMISC;
847 (*vlan_input_p)(ifp, m);
852 * Pass promiscuously received frames to the upper layer if the user
853 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
855 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
861 * Reset layer specific mbuf flags to avoid confusing upper layers.
862 * Strip off Ethernet header.
864 m->m_flags &= ~M_VLANTAG;
866 m_adj(m, ETHER_HDR_LEN);
869 * Dispatch frame to upper layer.
871 switch (ether_type) {
878 if (ifp->if_flags & IFF_NOARP) {
879 /* Discard packet if ARP is disabled on interface */
894 netisr_dispatch(isr, m);
899 * Packet is to be discarded. If netgraph is present,
900 * hand the packet to it for last chance processing;
901 * otherwise dispose of it.
903 if (ifp->if_l2com != NULL) {
904 KASSERT(ng_ether_input_orphan_p != NULL,
905 ("ng_ether_input_orphan_p is NULL"));
907 * Put back the ethernet header so netgraph has a
908 * consistent view of inbound packets.
910 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
911 (*ng_ether_input_orphan_p)(ifp, m);
918 * Convert Ethernet address to printable (loggable) representation.
919 * This routine is for compatibility; it's better to just use
921 * printf("%6D", <pointer to address>, ":");
923 * since there's no static buffer involved.
926 ether_sprintf(const u_char *ap)
928 static char etherbuf[18];
929 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
934 * Perform common duties while attaching to interface list
937 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
941 struct sockaddr_dl *sdl;
943 ifp->if_addrlen = ETHER_ADDR_LEN;
944 ifp->if_hdrlen = ETHER_HDR_LEN;
946 ifp->if_mtu = ETHERMTU;
947 ifp->if_output = ether_output;
948 ifp->if_input = ether_input;
949 ifp->if_resolvemulti = ether_resolvemulti;
950 ifp->if_requestencap = ether_requestencap;
952 ifp->if_reassign = ether_reassign;
954 if (ifp->if_baudrate == 0)
955 ifp->if_baudrate = IF_Mbps(10); /* just a default */
956 ifp->if_broadcastaddr = etherbroadcastaddr;
959 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
960 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
961 sdl->sdl_type = IFT_ETHER;
962 sdl->sdl_alen = ifp->if_addrlen;
963 bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
965 if (ifp->if_hw_addr != NULL)
966 bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen);
968 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
969 if (ng_ether_attach_p != NULL)
970 (*ng_ether_attach_p)(ifp);
972 /* Announce Ethernet MAC address if non-zero. */
973 for (i = 0; i < ifp->if_addrlen; i++)
976 if (i != ifp->if_addrlen)
977 if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
979 uuid_ether_add(LLADDR(sdl));
981 /* Add necessary bits are setup; announce it now. */
982 EVENTHANDLER_INVOKE(ether_ifattach_event, ifp);
983 if (IS_DEFAULT_VNET(curvnet))
984 devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL);
988 * Perform common duties while detaching an Ethernet interface
991 ether_ifdetach(struct ifnet *ifp)
993 struct sockaddr_dl *sdl;
995 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr);
996 uuid_ether_del(LLADDR(sdl));
998 if (ifp->if_l2com != NULL) {
999 KASSERT(ng_ether_detach_p != NULL,
1000 ("ng_ether_detach_p is NULL"));
1001 (*ng_ether_detach_p)(ifp);
1010 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused)
1013 if (ifp->if_l2com != NULL) {
1014 KASSERT(ng_ether_detach_p != NULL,
1015 ("ng_ether_detach_p is NULL"));
1016 (*ng_ether_detach_p)(ifp);
1019 if (ng_ether_attach_p != NULL) {
1020 CURVNET_SET_QUIET(new_vnet);
1021 (*ng_ether_attach_p)(ifp);
1027 SYSCTL_DECL(_net_link);
1028 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
1032 * This is for reference. We have a table-driven version
1033 * of the little-endian crc32 generator, which is faster
1034 * than the double-loop.
1037 ether_crc32_le(const uint8_t *buf, size_t len)
1044 crc = 0xffffffff; /* initial value */
1046 for (i = 0; i < len; i++) {
1047 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1048 carry = (crc ^ data) & 1;
1051 crc = (crc ^ ETHER_CRC_POLY_LE);
1059 ether_crc32_le(const uint8_t *buf, size_t len)
1061 static const uint32_t crctab[] = {
1062 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1063 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1064 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1065 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1070 crc = 0xffffffff; /* initial value */
1072 for (i = 0; i < len; i++) {
1074 crc = (crc >> 4) ^ crctab[crc & 0xf];
1075 crc = (crc >> 4) ^ crctab[crc & 0xf];
1083 ether_crc32_be(const uint8_t *buf, size_t len)
1086 uint32_t crc, carry;
1090 crc = 0xffffffff; /* initial value */
1092 for (i = 0; i < len; i++) {
1093 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1094 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1097 crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1105 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1107 struct ifaddr *ifa = (struct ifaddr *) data;
1108 struct ifreq *ifr = (struct ifreq *) data;
1113 ifp->if_flags |= IFF_UP;
1115 switch (ifa->ifa_addr->sa_family) {
1118 ifp->if_init(ifp->if_softc); /* before arpwhohas */
1119 arp_ifinit(ifp, ifa);
1123 ifp->if_init(ifp->if_softc);
1129 bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0],
1135 * Set the interface MTU.
1137 if (ifr->ifr_mtu > ETHERMTU) {
1140 ifp->if_mtu = ifr->ifr_mtu;
1145 error = priv_check(curthread, PRIV_NET_SETLANPCP);
1148 if (ifr->ifr_lan_pcp > 7 &&
1149 ifr->ifr_lan_pcp != IFNET_PCP_NONE) {
1152 ifp->if_pcp = ifr->ifr_lan_pcp;
1153 /* broadcast event about PCP change */
1154 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP);
1159 ifr->ifr_lan_pcp = ifp->if_pcp;
1163 error = EINVAL; /* XXX netbsd has ENOTTY??? */
1170 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1171 struct sockaddr *sa)
1173 struct sockaddr_dl *sdl;
1175 struct sockaddr_in *sin;
1178 struct sockaddr_in6 *sin6;
1182 switch(sa->sa_family) {
1185 * No mapping needed. Just check that it's a valid MC address.
1187 sdl = (struct sockaddr_dl *)sa;
1188 e_addr = LLADDR(sdl);
1189 if (!ETHER_IS_MULTICAST(e_addr))
1190 return EADDRNOTAVAIL;
1196 sin = (struct sockaddr_in *)sa;
1197 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1198 return EADDRNOTAVAIL;
1199 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1200 sdl->sdl_alen = ETHER_ADDR_LEN;
1201 e_addr = LLADDR(sdl);
1202 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1203 *llsa = (struct sockaddr *)sdl;
1208 sin6 = (struct sockaddr_in6 *)sa;
1209 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1211 * An IP6 address of 0 means listen to all
1212 * of the Ethernet multicast address used for IP6.
1213 * (This is used for multicast routers.)
1215 ifp->if_flags |= IFF_ALLMULTI;
1219 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1220 return EADDRNOTAVAIL;
1221 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1222 sdl->sdl_alen = ETHER_ADDR_LEN;
1223 e_addr = LLADDR(sdl);
1224 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1225 *llsa = (struct sockaddr *)sdl;
1231 * Well, the text isn't quite right, but it's the name
1234 return EAFNOSUPPORT;
1238 static moduledata_t ether_mod = {
1243 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1245 struct ether_vlan_header vlan;
1248 KASSERT((m->m_flags & M_VLANTAG) != 0,
1249 ("%s: vlan information not present", __func__));
1250 KASSERT(m->m_len >= sizeof(struct ether_header),
1251 ("%s: mbuf not large enough for header", __func__));
1252 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1253 vlan.evl_proto = vlan.evl_encap_proto;
1254 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1255 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1256 m->m_len -= sizeof(struct ether_header);
1257 m->m_data += sizeof(struct ether_header);
1259 * If a data link has been supplied by the caller, then we will need to
1260 * re-create a stack allocated mbuf chain with the following structure:
1262 * (1) mbuf #1 will contain the supplied data link
1263 * (2) mbuf #2 will contain the vlan header
1264 * (3) mbuf #3 will contain the original mbuf's packet data
1266 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1270 mv.m_data = (caddr_t)&vlan;
1271 mv.m_len = sizeof(vlan);
1277 bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1278 m->m_len += sizeof(struct ether_header);
1279 m->m_data -= sizeof(struct ether_header);
1283 ether_vlanencap(struct mbuf *m, uint16_t tag)
1285 struct ether_vlan_header *evl;
1287 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT);
1290 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1292 if (m->m_len < sizeof(*evl)) {
1293 m = m_pullup(m, sizeof(*evl));
1299 * Transform the Ethernet header into an Ethernet header
1300 * with 802.1Q encapsulation.
1302 evl = mtod(m, struct ether_vlan_header *);
1303 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1304 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1305 evl->evl_encap_proto = htons(ETHERTYPE_VLAN);
1306 evl->evl_tag = htons(tag);
1310 static SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW, 0,
1311 "IEEE 802.1Q VLAN");
1312 static SYSCTL_NODE(_net_link_vlan, PF_LINK, link, CTLFLAG_RW, 0,
1315 VNET_DEFINE_STATIC(int, soft_pad);
1316 #define V_soft_pad VNET(soft_pad)
1317 SYSCTL_INT(_net_link_vlan, OID_AUTO, soft_pad, CTLFLAG_RW | CTLFLAG_VNET,
1318 &VNET_NAME(soft_pad), 0,
1319 "pad short frames before tagging");
1322 * For now, make preserving PCP via an mbuf tag optional, as it increases
1323 * per-packet memory allocations and frees. In the future, it would be
1324 * preferable to reuse ether_vtag for this, or similar.
1326 int vlan_mtag_pcp = 0;
1327 SYSCTL_INT(_net_link_vlan, OID_AUTO, mtag_pcp, CTLFLAG_RW,
1329 "Retain VLAN PCP information as packets are passed up the stack");
1332 ether_8021q_frame(struct mbuf **mp, struct ifnet *ife, struct ifnet *p,
1333 uint16_t vid, uint8_t pcp)
1338 static const char pad[8]; /* just zeros */
1341 * Pad the frame to the minimum size allowed if told to.
1342 * This option is in accord with IEEE Std 802.1Q, 2003 Ed.,
1343 * paragraph C.4.4.3.b. It can help to work around buggy
1344 * bridges that violate paragraph C.4.4.3.a from the same
1345 * document, i.e., fail to pad short frames after untagging.
1346 * E.g., a tagged frame 66 bytes long (incl. FCS) is OK, but
1347 * untagging it will produce a 62-byte frame, which is a runt
1348 * and requires padding. There are VLAN-enabled network
1349 * devices that just discard such runts instead or mishandle
1352 if (V_soft_pad && p->if_type == IFT_ETHER) {
1353 for (n = ETHERMIN + ETHER_HDR_LEN - (*mp)->m_pkthdr.len;
1354 n > 0; n -= sizeof(pad)) {
1355 if (!m_append(*mp, min(n, sizeof(pad)), pad))
1361 if_printf(ife, "cannot pad short frame");
1367 * If underlying interface can do VLAN tag insertion itself,
1368 * just pass the packet along. However, we need some way to
1369 * tell the interface where the packet came from so that it
1370 * knows how to find the VLAN tag to use, so we attach a
1371 * packet tag that holds it.
1373 if (vlan_mtag_pcp && (mtag = m_tag_locate(*mp, MTAG_8021Q,
1374 MTAG_8021Q_PCP_OUT, NULL)) != NULL)
1375 tag = EVL_MAKETAG(vid, *(uint8_t *)(mtag + 1), 0);
1377 tag = EVL_MAKETAG(vid, pcp, 0);
1378 if (p->if_capenable & IFCAP_VLAN_HWTAGGING) {
1379 (*mp)->m_pkthdr.ether_vtag = tag;
1380 (*mp)->m_flags |= M_VLANTAG;
1382 *mp = ether_vlanencap(*mp, tag);
1384 if_printf(ife, "unable to prepend 802.1Q header");
1391 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1392 MODULE_VERSION(ether, 1);