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 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)if_ethersubr.c 8.1 (Berkeley) 6/10/93
34 * $FreeBSD: src/sys/net/if_ethersubr.c,v 1.70.2.33 2003/04/28 15:45:53 archie Exp $
35 * $DragonFly: src/sys/net/if_ethersubr.c,v 1.89 2008/09/17 08:51:29 sephe Exp $
38 #include "opt_atalk.h"
40 #include "opt_inet6.h"
43 #include "opt_netgraph.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/globaldata.h>
49 #include <sys/kernel.h>
50 #include <sys/malloc.h>
52 #include <sys/msgport.h>
53 #include <sys/socket.h>
54 #include <sys/sockio.h>
55 #include <sys/sysctl.h>
56 #include <sys/thread.h>
57 #include <sys/thread2.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>
65 #include <net/ifq_var.h>
67 #include <net/ethernet.h>
68 #include <net/vlan/if_vlan_ether.h>
69 #include <net/netmsg2.h>
71 #if defined(INET) || defined(INET6)
72 #include <netinet/in.h>
73 #include <netinet/in_var.h>
74 #include <netinet/if_ether.h>
75 #include <net/ipfw/ip_fw.h>
76 #include <net/dummynet/ip_dummynet.h>
79 #include <netinet6/nd6.h>
83 #include <netinet/ip_carp.h>
87 #include <netproto/ipx/ipx.h>
88 #include <netproto/ipx/ipx_if.h>
89 int (*ef_inputp)(struct ifnet*, const struct ether_header *eh, struct mbuf *m);
90 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst,
91 short *tp, int *hlen);
96 #include <netns/ns_if.h>
98 int ether_outputdebug = 0;
99 int ether_inputdebug = 0;
103 #include <netproto/atalk/at.h>
104 #include <netproto/atalk/at_var.h>
105 #include <netproto/atalk/at_extern.h>
107 #define llc_snap_org_code llc_un.type_snap.org_code
108 #define llc_snap_ether_type llc_un.type_snap.ether_type
110 extern u_char at_org_code[3];
111 extern u_char aarp_org_code[3];
112 #endif /* NETATALK */
115 #include <netproto/mpls/mpls.h>
118 /* netgraph node hooks for ng_ether(4) */
119 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
120 void (*ng_ether_input_orphan_p)(struct ifnet *ifp,
121 struct mbuf *m, const struct ether_header *eh);
122 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
123 void (*ng_ether_attach_p)(struct ifnet *ifp);
124 void (*ng_ether_detach_p)(struct ifnet *ifp);
126 void (*vlan_input_p)(struct mbuf *);
128 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
130 static void ether_restore_header(struct mbuf **, const struct ether_header *,
131 const struct ether_header *);
136 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
137 int (*bridge_output_p)(struct ifnet *, struct mbuf *);
138 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
140 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
143 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
144 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
147 #define gotoerr(e) do { error = (e); goto bad; } while (0)
148 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
150 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
152 const struct ether_header *eh);
154 static int ether_ipfw;
155 static u_int ether_restore_hdr;
156 static u_int ether_prepend_hdr;
158 SYSCTL_DECL(_net_link);
159 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
160 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
161 ðer_ipfw, 0, "Pass ether pkts through firewall");
162 SYSCTL_UINT(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW,
163 ðer_restore_hdr, 0, "# of ether header restoration");
164 SYSCTL_UINT(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW,
165 ðer_prepend_hdr, 0,
166 "# of ether header restoration which prepends mbuf");
169 * Ethernet output routine.
170 * Encapsulate a packet of type family for the local net.
171 * Use trailer local net encapsulation if enough data in first
172 * packet leaves a multiple of 512 bytes of data in remainder.
173 * Assumes that ifp is actually pointer to arpcom structure.
176 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
179 struct ether_header *eh, *deh;
182 int hlen = ETHER_HDR_LEN; /* link layer header length */
183 struct arpcom *ac = IFP2AC(ifp);
186 ASSERT_NOT_SERIALIZED(ifp->if_serializer);
188 if (ifp->if_flags & IFF_MONITOR)
190 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
193 M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT);
196 eh = mtod(m, struct ether_header *);
197 edst = eh->ether_dhost;
200 * Fill in the destination ethernet address and frame type.
202 switch (dst->sa_family) {
205 if (!arpresolve(ifp, rt, m, dst, edst))
206 return (0); /* if not yet resolved */
208 if (m->m_flags & M_MPLSLABELED)
209 eh->ether_type = htons(ETHERTYPE_MPLS);
212 eh->ether_type = htons(ETHERTYPE_IP);
217 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst))
218 return (0); /* Something bad happenned. */
219 eh->ether_type = htons(ETHERTYPE_IPV6);
224 if (ef_outputp != NULL) {
225 error = ef_outputp(ifp, &m, dst, &eh->ether_type,
230 eh->ether_type = htons(ETHERTYPE_IPX);
231 bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
232 edst, ETHER_ADDR_LEN);
238 struct at_ifaddr *aa;
240 if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL) {
245 * In the phase 2 case, need to prepend an mbuf for
246 * the llc header. Since we must preserve the value
247 * of m, which is passed to us by value, we m_copy()
248 * the first mbuf, and use it for our llc header.
250 if (aa->aa_flags & AFA_PHASE2) {
253 M_PREPEND(m, sizeof(struct llc), MB_DONTWAIT);
254 eh = mtod(m, struct ether_header *);
255 edst = eh->ether_dhost;
256 llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
257 llc.llc_control = LLC_UI;
258 bcopy(at_org_code, llc.llc_snap_org_code,
260 llc.llc_snap_ether_type = htons(ETHERTYPE_AT);
262 mtod(m, caddr_t) + sizeof(struct ether_header),
264 eh->ether_type = htons(m->m_pkthdr.len);
265 hlen = sizeof(struct llc) + ETHER_HDR_LEN;
267 eh->ether_type = htons(ETHERTYPE_AT);
269 if (!aarpresolve(ac, m, (struct sockaddr_at *)dst, edst))
278 case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
279 eh->ether_type = 0x8137;
281 case 0x0: /* Novell 802.3 */
282 eh->ether_type = htons(m->m_pkthdr.len);
284 case 0xe0e0: /* Novell 802.2 and Token-Ring */
285 M_PREPEND(m, 3, MB_DONTWAIT);
286 eh = mtod(m, struct ether_header *);
287 edst = eh->ether_dhost;
288 eh->ether_type = htons(m->m_pkthdr.len);
289 cp = mtod(m, u_char *) + sizeof(struct ether_header);
295 bcopy(&(((struct sockaddr_ns *)dst)->sns_addr.x_host), edst,
298 * XXX if ns_thishost is the same as the node's ethernet
299 * address then just the default code will catch this anyhow.
300 * So I'm not sure if this next clause should be here at all?
303 if (bcmp(edst, &ns_thishost, ETHER_ADDR_LEN) == 0) {
304 m->m_pkthdr.rcvif = ifp;
305 netisr_dispatch(NETISR_NS, m);
308 if (bcmp(edst, &ns_broadhost, ETHER_ADDR_LEN) == 0)
309 m->m_flags |= M_BCAST;
312 case pseudo_AF_HDRCMPLT:
314 loop_copy = -1; /* if this is for us, don't do it */
315 deh = (struct ether_header *)dst->sa_data;
316 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
317 eh->ether_type = deh->ether_type;
321 if_printf(ifp, "can't handle af%d\n", dst->sa_family);
322 gotoerr(EAFNOSUPPORT);
325 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */
326 memcpy(eh->ether_shost,
327 ((struct ether_header *)dst->sa_data)->ether_shost,
330 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
333 * Bridges require special output handling.
335 if (ifp->if_bridge) {
336 KASSERT(bridge_output_p != NULL,
337 ("%s: if_bridge not loaded!", __func__));
338 return bridge_output_p(ifp, m);
342 * If a simplex interface, and the packet is being sent to our
343 * Ethernet address or a broadcast address, loopback a copy.
344 * XXX To make a simplex device behave exactly like a duplex
345 * device, we should copy in the case of sending to our own
346 * ethernet address (thus letting the original actually appear
347 * on the wire). However, we don't do that here for security
348 * reasons and compatibility with the original behavior.
350 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
353 if (m->m_pkthdr.csum_flags & CSUM_IP)
354 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
355 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
356 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
357 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
360 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) {
361 n->m_pkthdr.csum_flags |= csum_flags;
362 if (csum_flags & CSUM_DATA_VALID)
363 n->m_pkthdr.csum_data = 0xffff;
364 if_simloop(ifp, n, dst->sa_family, hlen);
367 } else if (bcmp(eh->ether_dhost, eh->ether_shost,
368 ETHER_ADDR_LEN) == 0) {
369 m->m_pkthdr.csum_flags |= csum_flags;
370 if (csum_flags & CSUM_DATA_VALID)
371 m->m_pkthdr.csum_data = 0xffff;
372 if_simloop(ifp, m, dst->sa_family, hlen);
373 return (0); /* XXX */
378 if (ifp->if_carp && (error = carp_output(ifp, m, dst, NULL)))
383 /* Handle ng_ether(4) processing, if any */
384 if (ng_ether_output_p != NULL) {
385 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0)
391 /* Continue with link-layer output */
392 return ether_output_frame(ifp, m);
400 * Ethernet link layer output routine to send a raw frame to the device.
402 * This assumes that the 14 byte Ethernet header is present and contiguous
406 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
408 struct ip_fw *rule = NULL;
410 struct altq_pktattr pktattr;
412 ASSERT_NOT_SERIALIZED(ifp->if_serializer);
414 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
417 /* Extract info from dummynet tag */
418 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
419 KKASSERT(mtag != NULL);
420 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
421 KKASSERT(rule != NULL);
423 m_tag_delete(m, mtag);
424 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
427 if (ifq_is_enabled(&ifp->if_snd))
428 altq_etherclassify(&ifp->if_snd, m, &pktattr);
430 if (IPFW_LOADED && ether_ipfw != 0) {
431 struct ether_header save_eh, *eh;
433 eh = mtod(m, struct ether_header *);
435 m_adj(m, ETHER_HDR_LEN);
436 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) {
440 return ENOBUFS; /* pkt dropped */
442 return 0; /* consumed e.g. in a pipe */
445 /* packet was ok, restore the ethernet header */
446 ether_restore_header(&m, eh, &save_eh);
455 * Queue message on interface, update output statistics if
456 * successful, and start output if interface not yet active.
458 error = ifq_dispatch(ifp, m, &pktattr);
463 * ipfw processing for ethernet packets (in and out).
464 * The second parameter is NULL from ether_demux(), and ifp from
465 * ether_output_frame().
468 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule,
469 const struct ether_header *eh)
471 struct ether_header save_eh = *eh; /* might be a ptr in *m0 */
472 struct ip_fw_args args;
477 if (*rule != NULL && fw_one_pass)
478 return TRUE; /* dummynet packet, already partially processed */
481 * I need some amount of data to be contiguous.
483 i = min((*m0)->m_pkthdr.len, max_protohdr);
484 if ((*m0)->m_len < i) {
485 *m0 = m_pullup(*m0, i);
493 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
494 m_tag_delete(*m0, mtag);
495 if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
496 mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL);
497 KKASSERT(mtag != NULL);
498 m_tag_delete(*m0, mtag);
499 (*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED;
502 args.m = *m0; /* the packet we are looking at */
503 args.oif = dst; /* destination, if any */
504 args.rule = *rule; /* matching rule to restart */
505 args.eh = &save_eh; /* MAC header for bridged/MAC packets */
506 i = ip_fw_chk_ptr(&args);
521 * XXX at some point add support for divert/forward actions.
522 * If none of the above matches, we have to drop the pkt.
528 * Pass the pkt to dummynet, which consumes it.
530 m = *m0; /* pass the original to dummynet */
531 *m0 = NULL; /* and nothing back to the caller */
533 ether_restore_header(&m, eh, &save_eh);
537 ip_fw_dn_io_ptr(m, args.cookie,
538 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
543 panic("unknown ipfw return value: %d\n", i);
548 ether_input(struct ifnet *ifp, struct mbuf *m)
550 ether_input_chain(ifp, m, NULL);
554 * Perform common duties while attaching to interface list
557 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer)
559 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
564 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen,
565 lwkt_serialize_t serializer)
567 struct sockaddr_dl *sdl;
569 ifp->if_type = IFT_ETHER;
570 ifp->if_addrlen = ETHER_ADDR_LEN;
571 ifp->if_hdrlen = ETHER_HDR_LEN;
572 if_attach(ifp, serializer);
573 ifp->if_mtu = ETHERMTU;
574 if (ifp->if_baudrate == 0)
575 ifp->if_baudrate = 10000000;
576 ifp->if_output = ether_output;
577 ifp->if_input = ether_input;
578 ifp->if_resolvemulti = ether_resolvemulti;
579 ifp->if_broadcastaddr = etherbroadcastaddr;
580 sdl = IF_LLSOCKADDR(ifp);
581 sdl->sdl_type = IFT_ETHER;
582 sdl->sdl_alen = ifp->if_addrlen;
583 bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
585 * XXX Keep the current drivers happy.
586 * XXX Remove once all drivers have been cleaned up
588 if (lla != IFP2AC(ifp)->ac_enaddr)
589 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
590 bpfattach(ifp, dlt, hdrlen);
591 if (ng_ether_attach_p != NULL)
592 (*ng_ether_attach_p)(ifp);
594 if_printf(ifp, "MAC address: %6D\n", lla, ":");
598 * Perform common duties while detaching an Ethernet interface
601 ether_ifdetach(struct ifnet *ifp)
605 if (ng_ether_detach_p != NULL)
606 (*ng_ether_detach_p)(ifp);
612 ether_ioctl(struct ifnet *ifp, int command, caddr_t data)
614 struct ifaddr *ifa = (struct ifaddr *) data;
615 struct ifreq *ifr = (struct ifreq *) data;
618 #define IF_INIT(ifp) \
620 if (((ifp)->if_flags & IFF_UP) == 0) { \
621 (ifp)->if_flags |= IFF_UP; \
622 (ifp)->if_init((ifp)->if_softc); \
626 ASSERT_SERIALIZED(ifp->if_serializer);
630 switch (ifa->ifa_addr->sa_family) {
633 IF_INIT(ifp); /* before arpwhohas */
634 arp_ifinit(ifp, ifa);
639 * XXX - This code is probably wrong
643 struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr;
644 struct arpcom *ac = IFP2AC(ifp);
646 if (ipx_nullhost(*ina))
647 ina->x_host = *(union ipx_host *) ac->ac_enaddr;
649 bcopy(ina->x_host.c_host, ac->ac_enaddr,
650 sizeof ac->ac_enaddr);
652 IF_INIT(ifp); /* Set new address. */
658 * XXX - This code is probably wrong
662 struct ns_addr *ina = &(IA_SNS(ifa)->sns_addr);
663 struct arpcom *ac = IFP2AC(ifp);
665 if (ns_nullhost(*ina))
666 ina->x_host = *(union ns_host *)(ac->ac_enaddr);
668 bcopy(ina->x_host.c_host, ac->ac_enaddr,
669 sizeof ac->ac_enaddr);
685 bcopy(IFP2AC(ifp)->ac_enaddr,
686 ((struct sockaddr *)ifr->ifr_data)->sa_data,
692 * Set the interface MTU.
694 if (ifr->ifr_mtu > ETHERMTU) {
697 ifp->if_mtu = ifr->ifr_mtu;
712 struct sockaddr **llsa,
715 struct sockaddr_dl *sdl;
716 struct sockaddr_in *sin;
718 struct sockaddr_in6 *sin6;
722 switch(sa->sa_family) {
725 * No mapping needed. Just check that it's a valid MC address.
727 sdl = (struct sockaddr_dl *)sa;
728 e_addr = LLADDR(sdl);
729 if ((e_addr[0] & 1) != 1)
730 return EADDRNOTAVAIL;
736 sin = (struct sockaddr_in *)sa;
737 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
738 return EADDRNOTAVAIL;
739 MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
741 sdl->sdl_len = sizeof *sdl;
742 sdl->sdl_family = AF_LINK;
743 sdl->sdl_index = ifp->if_index;
744 sdl->sdl_type = IFT_ETHER;
745 sdl->sdl_alen = ETHER_ADDR_LEN;
746 e_addr = LLADDR(sdl);
747 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
748 *llsa = (struct sockaddr *)sdl;
753 sin6 = (struct sockaddr_in6 *)sa;
754 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
756 * An IP6 address of 0 means listen to all
757 * of the Ethernet multicast address used for IP6.
758 * (This is used for multicast routers.)
760 ifp->if_flags |= IFF_ALLMULTI;
764 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
765 return EADDRNOTAVAIL;
766 MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
768 sdl->sdl_len = sizeof *sdl;
769 sdl->sdl_family = AF_LINK;
770 sdl->sdl_index = ifp->if_index;
771 sdl->sdl_type = IFT_ETHER;
772 sdl->sdl_alen = ETHER_ADDR_LEN;
773 e_addr = LLADDR(sdl);
774 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
775 *llsa = (struct sockaddr *)sdl;
781 * Well, the text isn't quite right, but it's the name
790 * This is for reference. We have a table-driven version
791 * of the little-endian crc32 generator, which is faster
792 * than the double-loop.
795 ether_crc32_le(const uint8_t *buf, size_t len)
797 uint32_t c, crc, carry;
800 crc = 0xffffffffU; /* initial value */
802 for (i = 0; i < len; i++) {
804 for (j = 0; j < 8; j++) {
805 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
809 crc = (crc ^ ETHER_CRC_POLY_LE);
817 ether_crc32_le(const uint8_t *buf, size_t len)
819 static const uint32_t crctab[] = {
820 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
821 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
822 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
823 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
828 crc = 0xffffffffU; /* initial value */
830 for (i = 0; i < len; i++) {
832 crc = (crc >> 4) ^ crctab[crc & 0xf];
833 crc = (crc >> 4) ^ crctab[crc & 0xf];
841 ether_crc32_be(const uint8_t *buf, size_t len)
843 uint32_t c, crc, carry;
846 crc = 0xffffffffU; /* initial value */
848 for (i = 0; i < len; i++) {
850 for (j = 0; j < 8; j++) {
851 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
855 crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
863 * find the size of ethernet header, and call classifier
866 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
867 struct altq_pktattr *pktattr)
869 struct ether_header *eh;
871 int hlen, af, hdrsize;
874 hlen = sizeof(struct ether_header);
875 eh = mtod(m, struct ether_header *);
877 ether_type = ntohs(eh->ether_type);
878 if (ether_type < ETHERMTU) {
880 struct llc *llc = (struct llc *)(eh + 1);
883 if (m->m_len < hlen ||
884 llc->llc_dsap != LLC_SNAP_LSAP ||
885 llc->llc_ssap != LLC_SNAP_LSAP ||
886 llc->llc_control != LLC_UI)
887 goto bad; /* not snap! */
889 ether_type = ntohs(llc->llc_un.type_snap.ether_type);
892 if (ether_type == ETHERTYPE_IP) {
894 hdrsize = 20; /* sizeof(struct ip) */
896 } else if (ether_type == ETHERTYPE_IPV6) {
898 hdrsize = 40; /* sizeof(struct ip6_hdr) */
903 while (m->m_len <= hlen) {
907 hdr = m->m_data + hlen;
908 if (m->m_len < hlen + hdrsize) {
910 * ip header is not in a single mbuf. this should not
911 * happen in the current code.
912 * (todo: use m_pulldown in the future)
918 ifq_classify(ifq, m, af, pktattr);
925 pktattr->pattr_class = NULL;
926 pktattr->pattr_hdr = NULL;
927 pktattr->pattr_af = AF_UNSPEC;
931 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
932 const struct ether_header *save_eh)
934 struct mbuf *m = *m0;
939 * Prepend the header, optimize for the common case of
940 * eh pointing into the mbuf.
942 if ((const void *)(eh + 1) == (void *)m->m_data) {
943 m->m_data -= ETHER_HDR_LEN;
944 m->m_len += ETHER_HDR_LEN;
945 m->m_pkthdr.len += ETHER_HDR_LEN;
949 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
951 bcopy(save_eh, mtod(m, struct ether_header *),
959 ether_input_ipifunc(void *arg)
961 struct mbuf *m, *next;
969 port = m->m_pkthdr.header;
970 m->m_pkthdr.header = NULL;
973 &m->m_hdr.mh_netmsg.nm_netmsg.nm_lmsg);
980 ether_input_dispatch(struct mbuf_chain *chain)
985 for (i = 0; i < ncpus; ++i) {
986 if (chain[i].mc_head != NULL) {
987 lwkt_send_ipiq(globaldata_find(i),
988 ether_input_ipifunc, chain[i].mc_head);
992 if (chain->mc_head != NULL)
993 ether_input_ipifunc(chain->mc_head);
998 ether_input_chain_init(struct mbuf_chain *chain)
1003 for (i = 0; i < ncpus; ++i)
1004 chain[i].mc_head = chain[i].mc_tail = NULL;
1006 chain->mc_head = chain->mc_tail = NULL;
1011 * Upper layer processing for a received Ethernet packet.
1014 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
1016 struct ether_header *eh;
1017 int isr, redispatch;
1019 struct ip_fw *rule = NULL;
1025 KASSERT(m->m_len >= ETHER_HDR_LEN,
1026 ("ether header is no contiguous!\n"));
1028 eh = mtod(m, struct ether_header *);
1030 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
1033 /* Extract info from dummynet tag */
1034 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
1035 KKASSERT(mtag != NULL);
1036 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
1037 KKASSERT(rule != NULL);
1039 m_tag_delete(m, mtag);
1040 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
1042 /* packet is passing the second time */
1048 * XXX: Okay, we need to call carp_forus() and - if it is for
1049 * us jump over code that does the normal check
1050 * "ac_enaddr == ether_dhost". The check sequence is a bit
1051 * different from OpenBSD, so we jump over as few code as
1052 * possible, to catch _all_ sanity checks. This needs
1053 * evaluation, to see if the carp ether_dhost values break any
1056 if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost))
1061 * Discard packet if upper layers shouldn't see it because
1062 * it was unicast to a different Ethernet address. If the
1063 * driver is working properly, then this situation can only
1064 * happen when the interface is in promiscuous mode.
1066 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
1067 (eh->ether_dhost[0] & 1) == 0 &&
1068 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
1074 if (IPFW_LOADED && ether_ipfw != 0) {
1075 struct ether_header save_eh = *eh;
1077 /* XXX old crufty stuff, needs to be removed */
1078 m_adj(m, sizeof(struct ether_header));
1080 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
1085 ether_restore_header(&m, eh, &save_eh);
1088 eh = mtod(m, struct ether_header *);
1091 ether_type = ntohs(eh->ether_type);
1092 KKASSERT(ether_type != ETHERTYPE_VLAN);
1094 if (m->m_flags & M_VLANTAG) {
1095 if (vlan_input_p != NULL) {
1098 m->m_pkthdr.rcvif->if_noproto++;
1104 m_adj(m, sizeof(struct ether_header));
1107 switch (ether_type) {
1110 if (ipflow_fastforward(m))
1116 if (ifp->if_flags & IFF_NOARP) {
1117 /* Discard packet if ARP is disabled on interface */
1126 case ETHERTYPE_IPV6:
1133 if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
1140 case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
1148 isr = NETISR_ATALK1;
1150 case ETHERTYPE_AARP:
1156 case ETHERTYPE_MPLS:
1157 case ETHERTYPE_MPLS_MCAST:
1158 /* Should have been set by ether_input_chain(). */
1159 KKASSERT(m->m_flags & M_MPLSLABELED);
1166 * The accurate msgport is not determined before
1167 * we reach here, so redo the dispatching
1171 if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
1175 checksum = mtod(m, ushort *);
1177 if ((ether_type <= ETHERMTU) &&
1178 ((*checksum == 0xffff) || (*checksum == 0xE0E0))) {
1179 if (*checksum == 0xE0E0) {
1180 m->m_pkthdr.len -= 3;
1189 if (ether_type > ETHERMTU)
1191 l = mtod(m, struct llc *);
1192 if (l->llc_dsap == LLC_SNAP_LSAP &&
1193 l->llc_ssap == LLC_SNAP_LSAP &&
1194 l->llc_control == LLC_UI) {
1195 if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
1196 sizeof at_org_code) == 0 &&
1197 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
1198 m_adj(m, sizeof(struct llc));
1199 isr = NETISR_ATALK2;
1202 if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
1203 sizeof aarp_org_code) == 0 &&
1204 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
1205 m_adj(m, sizeof(struct llc));
1212 if (ng_ether_input_orphan_p != NULL)
1213 ng_ether_input_orphan_p(ifp, m, eh);
1222 netisr_dispatch(isr, m);
1226 * First we perform any link layer operations, then continue to the
1227 * upper layers with ether_demux_oncpu().
1230 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1232 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1234 * Receiving interface's flags are changed, when this
1235 * packet is waiting for processing; discard it.
1242 * Tap the packet off here for a bridge. bridge_input()
1243 * will return NULL if it has consumed the packet, otherwise
1244 * it gets processed as normal. Note that bridge_input()
1245 * will always return the original packet if we need to
1246 * process it locally.
1248 if (ifp->if_bridge) {
1249 KASSERT(bridge_input_p != NULL,
1250 ("%s: if_bridge not loaded!", __func__));
1252 if(m->m_flags & M_PROTO1) {
1253 m->m_flags &= ~M_PROTO1;
1255 /* clear M_PROMISC, in case the packets comes from a vlan */
1256 /* m->m_flags &= ~M_PROMISC; */
1257 m = bridge_input_p(ifp, m);
1261 KASSERT(ifp == m->m_pkthdr.rcvif,
1262 ("bridge_input_p changed rcvif\n"));
1266 /* Handle ng_ether(4) processing, if any */
1267 if (ng_ether_input_p != NULL) {
1268 ng_ether_input_p(ifp, &m);
1273 /* Continue with upper layer processing */
1274 ether_demux_oncpu(ifp, m);
1278 ether_input_handler(struct netmsg *nmsg)
1280 struct netmsg_packet *nmp = (struct netmsg_packet *)nmsg;
1286 ifp = m->m_pkthdr.rcvif;
1288 ether_input_oncpu(ifp, m);
1291 static __inline void
1292 ether_init_netpacket(int num, struct mbuf *m)
1294 struct netmsg_packet *pmsg;
1296 pmsg = &m->m_hdr.mh_netmsg;
1297 netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport, 0,
1298 ether_input_handler);
1299 pmsg->nm_packet = m;
1300 pmsg->nm_netmsg.nm_lmsg.u.ms_result = num;
1303 static __inline struct lwkt_port *
1304 ether_mport(int num, struct mbuf **m)
1306 if (num == NETISR_MAX) {
1308 * All packets whose target msgports can't be
1309 * determined here are dispatched to netisr0,
1310 * where further dispatching may happen.
1312 return cpu_portfn(0);
1314 return netisr_find_port(num, m);
1318 * Process a received Ethernet packet.
1320 * The ethernet header is assumed to be in the mbuf so the caller
1321 * MUST MAKE SURE that there are at least sizeof(struct ether_header)
1322 * bytes in the first mbuf.
1324 * We first try to find the target msgport for this ether frame, if
1325 * there is no target msgport for it, this ether frame is discarded,
1326 * else we do following processing according to whether 'chain' is
1328 * - If 'chain' is NULL, this ether frame is sent to the target msgport
1329 * immediately. This situation happens when ether_input_chain is
1330 * accessed through ifnet.if_input.
1331 * - If 'chain' is not NULL, this ether frame is queued to the 'chain'
1332 * bucket indexed by the target msgport's cpuid and the target msgport
1333 * is saved in mbuf's m_pkthdr.m_head. Caller of ether_input_chain
1334 * must initialize 'chain' by calling ether_input_chain_init().
1335 * ether_input_dispatch must be called later to send ether frames
1336 * queued on 'chain' to their target msgport.
1339 ether_input_chain(struct ifnet *ifp, struct mbuf *m, struct mbuf_chain *chain)
1341 struct ether_header *eh, *save_eh, save_eh0;
1342 struct lwkt_port *port;
1343 uint16_t ether_type;
1346 ASSERT_SERIALIZED(ifp->if_serializer);
1349 /* Discard packet if interface is not up */
1350 if (!(ifp->if_flags & IFF_UP)) {
1355 if (m->m_len < sizeof(struct ether_header)) {
1356 /* XXX error in the caller. */
1360 eh = mtod(m, struct ether_header *);
1362 m->m_pkthdr.rcvif = ifp;
1364 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1365 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1366 ifp->if_addrlen) == 0)
1367 m->m_flags |= M_BCAST;
1369 m->m_flags |= M_MCAST;
1373 ETHER_BPF_MTAP(ifp, m);
1375 ifp->if_ibytes += m->m_pkthdr.len;
1377 if (ifp->if_flags & IFF_MONITOR) {
1379 * Interface marked for monitoring; discard packet.
1385 if (ntohs(eh->ether_type) == ETHERTYPE_VLAN &&
1386 (m->m_flags & M_VLANTAG) == 0) {
1388 * Extract vlan tag if hardware does not do it for us
1390 vlan_ether_decap(&m);
1393 eh = mtod(m, struct ether_header *);
1395 ether_type = ntohs(eh->ether_type);
1397 if ((m->m_flags & M_VLANTAG) && ether_type == ETHERTYPE_VLAN) {
1399 * To prevent possible dangerous recursion,
1400 * we don't do vlan-in-vlan
1406 KKASSERT(ether_type != ETHERTYPE_VLAN);
1409 * Map ether type to netisr id.
1411 switch (ether_type) {
1423 case ETHERTYPE_IPV6:
1435 case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
1442 isr = NETISR_ATALK1;
1444 case ETHERTYPE_AARP:
1450 case ETHERTYPE_MPLS:
1451 case ETHERTYPE_MPLS_MCAST:
1452 m->m_flags |= M_MPLSLABELED;
1459 * NETISR_MAX is an invalid value; it is chosen to let
1460 * ether_mport() know that we are not able to decide
1461 * this packet's msgport here.
1468 * If the packet is in contiguous memory, following
1469 * m_adj() could ensure that the hidden ether header
1470 * will not be destroyed, else we will have to save
1471 * the ether header for the later restoration.
1473 if (m->m_pkthdr.len != m->m_len) {
1475 save_eh = &save_eh0;
1481 * Temporarily remove ether header; ether_mport()
1482 * expects a packet without ether header.
1484 m_adj(m, sizeof(struct ether_header));
1487 * Find the packet's target msgport.
1489 port = ether_mport(isr, &m);
1491 KKASSERT(m == NULL);
1496 * Restore ether header.
1498 if (save_eh != NULL) {
1499 ether_restore_header(&m, eh, save_eh);
1503 m->m_data -= ETHER_HDR_LEN;
1504 m->m_len += ETHER_HDR_LEN;
1505 m->m_pkthdr.len += ETHER_HDR_LEN;
1509 * Initialize mbuf's netmsg packet _after_ possible
1510 * ether header restoration, else the initialized
1511 * netmsg packet may be lost during ether header
1514 ether_init_netpacket(isr, m);
1516 if (chain != NULL) {
1517 struct mbuf_chain *c;
1520 m->m_pkthdr.header = port; /* XXX */
1521 cpuid = port->mpu_td->td_gd->gd_cpuid;
1524 if (c->mc_head == NULL) {
1525 c->mc_head = c->mc_tail = m;
1527 c->mc_tail->m_nextpkt = m;
1530 m->m_nextpkt = NULL;
1532 lwkt_sendmsg(port, &m->m_hdr.mh_netmsg.nm_netmsg.nm_lmsg);