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 $
38 #include "opt_inet6.h"
41 #include "opt_netgraph.h"
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/globaldata.h>
48 #include <sys/kernel.h>
51 #include <sys/malloc.h>
53 #include <sys/msgport.h>
54 #include <sys/socket.h>
55 #include <sys/sockio.h>
56 #include <sys/sysctl.h>
57 #include <sys/thread.h>
59 #include <sys/thread2.h>
60 #include <sys/mplock2.h>
63 #include <net/netisr.h>
64 #include <net/route.h>
65 #include <net/if_llc.h>
66 #include <net/if_dl.h>
67 #include <net/if_types.h>
68 #include <net/ifq_var.h>
70 #include <net/ethernet.h>
71 #include <net/vlan/if_vlan_ether.h>
72 #include <net/vlan/if_vlan_var.h>
73 #include <net/netmsg2.h>
75 #if defined(INET) || defined(INET6)
76 #include <netinet/in.h>
77 #include <netinet/ip_var.h>
78 #include <netinet/tcp_var.h>
79 #include <netinet/if_ether.h>
80 #include <netinet/ip_flow.h>
81 #include <net/ipfw/ip_fw.h>
82 #include <net/dummynet/ip_dummynet.h>
85 #include <netinet6/nd6.h>
89 #include <netinet/ip_carp.h>
93 #include <netproto/ipx/ipx.h>
94 #include <netproto/ipx/ipx_if.h>
95 int (*ef_inputp)(struct ifnet*, const struct ether_header *eh, struct mbuf *m);
96 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst,
97 short *tp, int *hlen);
101 #include <netproto/mpls/mpls.h>
104 /* netgraph node hooks for ng_ether(4) */
105 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
106 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
107 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
108 void (*ng_ether_attach_p)(struct ifnet *ifp);
109 void (*ng_ether_detach_p)(struct ifnet *ifp);
111 void (*vlan_input_p)(struct mbuf *);
113 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
115 static void ether_restore_header(struct mbuf **, const struct ether_header *,
116 const struct ether_header *);
117 static int ether_characterize(struct mbuf **);
122 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
123 int (*bridge_output_p)(struct ifnet *, struct mbuf *);
124 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
125 struct ifnet *(*bridge_interface_p)(void *if_bridge);
127 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
130 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
131 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
134 #define gotoerr(e) do { error = (e); goto bad; } while (0)
135 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
137 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
139 const struct ether_header *eh);
141 static int ether_ipfw;
142 static u_long ether_restore_hdr;
143 static u_long ether_prepend_hdr;
144 static u_long ether_input_wronghash;
145 static int ether_debug;
148 static u_long ether_pktinfo_try;
149 static u_long ether_pktinfo_hit;
150 static u_long ether_rss_nopi;
151 static u_long ether_rss_nohash;
152 static u_long ether_input_requeue;
155 SYSCTL_DECL(_net_link);
156 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
157 SYSCTL_INT(_net_link_ether, OID_AUTO, debug, CTLFLAG_RW,
158 ðer_debug, 0, "Ether debug");
159 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
160 ðer_ipfw, 0, "Pass ether pkts through firewall");
161 SYSCTL_ULONG(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW,
162 ðer_restore_hdr, 0, "# of ether header restoration");
163 SYSCTL_ULONG(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW,
164 ðer_prepend_hdr, 0,
165 "# of ether header restoration which prepends mbuf");
166 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghash, CTLFLAG_RW,
167 ðer_input_wronghash, 0, "# of input packets with wrong hash");
169 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nopi, CTLFLAG_RW,
170 ðer_rss_nopi, 0, "# of packets do not have pktinfo");
171 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nohash, CTLFLAG_RW,
172 ðer_rss_nohash, 0, "# of packets do not have hash");
173 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_try, CTLFLAG_RW,
174 ðer_pktinfo_try, 0,
175 "# of tries to find packets' msgport using pktinfo");
176 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_hit, CTLFLAG_RW,
177 ðer_pktinfo_hit, 0,
178 "# of packets whose msgport are found using pktinfo");
179 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_requeue, CTLFLAG_RW,
180 ðer_input_requeue, 0, "# of input packets gets requeued");
183 #define ETHER_KTR_STR "ifp=%p"
184 #define ETHER_KTR_ARGS struct ifnet *ifp
186 #define KTR_ETHERNET KTR_ALL
188 KTR_INFO_MASTER(ether);
189 KTR_INFO(KTR_ETHERNET, ether, pkt_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARGS);
190 KTR_INFO(KTR_ETHERNET, ether, pkt_end, 1, ETHER_KTR_STR, ETHER_KTR_ARGS);
191 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARGS);
192 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARGS);
193 #define logether(name, arg) KTR_LOG(ether_ ## name, arg)
196 * Ethernet output routine.
197 * Encapsulate a packet of type family for the local net.
198 * Use trailer local net encapsulation if enough data in first
199 * packet leaves a multiple of 512 bytes of data in remainder.
200 * Assumes that ifp is actually pointer to arpcom structure.
203 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
206 struct ether_header *eh, *deh;
209 int hlen = ETHER_HDR_LEN; /* link layer header length */
210 struct arpcom *ac = IFP2AC(ifp);
213 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
215 if (ifp->if_flags & IFF_MONITOR)
217 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
220 M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT);
223 m->m_pkthdr.csum_lhlen = sizeof(struct ether_header);
224 eh = mtod(m, struct ether_header *);
225 edst = eh->ether_dhost;
228 * Fill in the destination ethernet address and frame type.
230 switch (dst->sa_family) {
233 if (!arpresolve(ifp, rt, m, dst, edst))
234 return (0); /* if not yet resolved */
236 if (m->m_flags & M_MPLSLABELED)
237 eh->ether_type = htons(ETHERTYPE_MPLS);
240 eh->ether_type = htons(ETHERTYPE_IP);
245 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst))
246 return (0); /* Something bad happenned. */
247 eh->ether_type = htons(ETHERTYPE_IPV6);
252 if (ef_outputp != NULL) {
254 * Hold BGL and recheck ef_outputp
257 if (ef_outputp != NULL) {
258 error = ef_outputp(ifp, &m, dst,
259 &eh->ether_type, &hlen);
268 eh->ether_type = htons(ETHERTYPE_IPX);
269 bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
270 edst, ETHER_ADDR_LEN);
273 case pseudo_AF_HDRCMPLT:
275 loop_copy = -1; /* if this is for us, don't do it */
276 deh = (struct ether_header *)dst->sa_data;
277 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
278 eh->ether_type = deh->ether_type;
282 if_printf(ifp, "can't handle af%d\n", dst->sa_family);
283 gotoerr(EAFNOSUPPORT);
286 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */
287 memcpy(eh->ether_shost,
288 ((struct ether_header *)dst->sa_data)->ether_shost,
291 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
294 * Bridges require special output handling.
296 if (ifp->if_bridge) {
297 KASSERT(bridge_output_p != NULL,
298 ("%s: if_bridge not loaded!", __func__));
299 return bridge_output_p(ifp, m);
303 * If a simplex interface, and the packet is being sent to our
304 * Ethernet address or a broadcast address, loopback a copy.
305 * XXX To make a simplex device behave exactly like a duplex
306 * device, we should copy in the case of sending to our own
307 * ethernet address (thus letting the original actually appear
308 * on the wire). However, we don't do that here for security
309 * reasons and compatibility with the original behavior.
311 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
314 if (m->m_pkthdr.csum_flags & CSUM_IP)
315 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
316 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
317 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
318 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
321 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) {
322 n->m_pkthdr.csum_flags |= csum_flags;
323 if (csum_flags & CSUM_DATA_VALID)
324 n->m_pkthdr.csum_data = 0xffff;
325 if_simloop(ifp, n, dst->sa_family, hlen);
328 } else if (bcmp(eh->ether_dhost, eh->ether_shost,
329 ETHER_ADDR_LEN) == 0) {
330 m->m_pkthdr.csum_flags |= csum_flags;
331 if (csum_flags & CSUM_DATA_VALID)
332 m->m_pkthdr.csum_data = 0xffff;
333 if_simloop(ifp, m, dst->sa_family, hlen);
334 return (0); /* XXX */
339 if (ifp->if_type == IFT_CARP) {
340 ifp = carp_parent(ifp);
342 gotoerr(ENETUNREACH);
347 * Check precondition again
349 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
351 if (ifp->if_flags & IFF_MONITOR)
353 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
354 (IFF_UP | IFF_RUNNING))
359 /* Handle ng_ether(4) processing, if any */
360 if (ng_ether_output_p != NULL) {
362 * Hold BGL and recheck ng_ether_output_p
365 if (ng_ether_output_p != NULL) {
366 if ((error = ng_ether_output_p(ifp, &m)) != 0) {
378 /* Continue with link-layer output */
379 return ether_output_frame(ifp, m);
387 * Returns the bridge interface an ifp is associated
390 * Only call if ifp->if_bridge != NULL.
393 ether_bridge_interface(struct ifnet *ifp)
395 if (bridge_interface_p)
396 return(bridge_interface_p(ifp->if_bridge));
401 * Ethernet link layer output routine to send a raw frame to the device.
403 * This assumes that the 14 byte Ethernet header is present and contiguous
407 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
409 struct ip_fw *rule = NULL;
411 struct altq_pktattr pktattr;
413 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
415 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
418 /* Extract info from dummynet tag */
419 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
420 KKASSERT(mtag != NULL);
421 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
422 KKASSERT(rule != NULL);
424 m_tag_delete(m, mtag);
425 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
428 if (ifq_is_enabled(&ifp->if_snd))
429 altq_etherclassify(&ifp->if_snd, m, &pktattr);
431 if (IPFW_LOADED && ether_ipfw != 0) {
432 struct ether_header save_eh, *eh;
434 eh = mtod(m, struct ether_header *);
436 m_adj(m, ETHER_HDR_LEN);
437 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) {
441 return ENOBUFS; /* pkt dropped */
443 return 0; /* consumed e.g. in a pipe */
446 /* packet was ok, restore the ethernet header */
447 ether_restore_header(&m, eh, &save_eh);
456 * Queue message on interface, update output statistics if
457 * successful, and start output if interface not yet active.
459 error = ifq_dispatch(ifp, m, &pktattr);
464 * ipfw processing for ethernet packets (in and out).
465 * The second parameter is NULL from ether_demux(), and ifp from
466 * ether_output_frame().
469 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule,
470 const struct ether_header *eh)
472 struct ether_header save_eh = *eh; /* might be a ptr in *m0 */
473 struct ip_fw_args args;
478 if (*rule != NULL && fw_one_pass)
479 return TRUE; /* dummynet packet, already partially processed */
482 * I need some amount of data to be contiguous.
484 i = min((*m0)->m_pkthdr.len, max_protohdr);
485 if ((*m0)->m_len < i) {
486 *m0 = m_pullup(*m0, i);
494 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
495 m_tag_delete(*m0, mtag);
496 if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
497 mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL);
498 KKASSERT(mtag != NULL);
499 m_tag_delete(*m0, mtag);
500 (*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED;
503 args.m = *m0; /* the packet we are looking at */
504 args.oif = dst; /* destination, if any */
505 args.rule = *rule; /* matching rule to restart */
506 args.eh = &save_eh; /* MAC header for bridged/MAC packets */
507 i = ip_fw_chk_ptr(&args);
522 * XXX at some point add support for divert/forward actions.
523 * If none of the above matches, we have to drop the pkt.
529 * Pass the pkt to dummynet, which consumes it.
531 m = *m0; /* pass the original to dummynet */
532 *m0 = NULL; /* and nothing back to the caller */
534 ether_restore_header(&m, eh, &save_eh);
538 ip_fw_dn_io_ptr(m, args.cookie,
539 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
544 panic("unknown ipfw return value: %d", i);
549 ether_input(struct ifnet *ifp, struct mbuf *m)
551 ether_input_pkt(ifp, m, NULL);
555 * Perform common duties while attaching to interface list
558 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer)
560 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
565 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen,
566 lwkt_serialize_t serializer)
568 struct sockaddr_dl *sdl;
570 ifp->if_type = IFT_ETHER;
571 ifp->if_addrlen = ETHER_ADDR_LEN;
572 ifp->if_hdrlen = ETHER_HDR_LEN;
573 if_attach(ifp, serializer);
574 ifp->if_mtu = ETHERMTU;
575 if (ifp->if_baudrate == 0)
576 ifp->if_baudrate = 10000000;
577 ifp->if_output = ether_output;
578 ifp->if_input = ether_input;
579 ifp->if_resolvemulti = ether_resolvemulti;
580 ifp->if_broadcastaddr = etherbroadcastaddr;
581 sdl = IF_LLSOCKADDR(ifp);
582 sdl->sdl_type = IFT_ETHER;
583 sdl->sdl_alen = ifp->if_addrlen;
584 bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
586 * XXX Keep the current drivers happy.
587 * XXX Remove once all drivers have been cleaned up
589 if (lla != IFP2AC(ifp)->ac_enaddr)
590 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
591 bpfattach(ifp, dlt, hdrlen);
592 if (ng_ether_attach_p != NULL)
593 (*ng_ether_attach_p)(ifp);
595 if_printf(ifp, "MAC address: %6D\n", lla, ":");
599 * Perform common duties while detaching an Ethernet interface
602 ether_ifdetach(struct ifnet *ifp)
606 if (ng_ether_detach_p != NULL)
607 (*ng_ether_detach_p)(ifp);
613 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
615 struct ifaddr *ifa = (struct ifaddr *) data;
616 struct ifreq *ifr = (struct ifreq *) data;
619 #define IF_INIT(ifp) \
621 if (((ifp)->if_flags & IFF_UP) == 0) { \
622 (ifp)->if_flags |= IFF_UP; \
623 (ifp)->if_init((ifp)->if_softc); \
627 ASSERT_IFNET_SERIALIZED_ALL(ifp);
631 switch (ifa->ifa_addr->sa_family) {
634 IF_INIT(ifp); /* before arpwhohas */
635 arp_ifinit(ifp, ifa);
640 * XXX - This code is probably wrong
644 struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr;
645 struct arpcom *ac = IFP2AC(ifp);
647 if (ipx_nullhost(*ina))
648 ina->x_host = *(union ipx_host *) ac->ac_enaddr;
650 bcopy(ina->x_host.c_host, ac->ac_enaddr,
651 sizeof ac->ac_enaddr);
653 IF_INIT(ifp); /* Set new address. */
664 bcopy(IFP2AC(ifp)->ac_enaddr,
665 ((struct sockaddr *)ifr->ifr_data)->sa_data,
671 * Set the interface MTU.
673 if (ifr->ifr_mtu > ETHERMTU) {
676 ifp->if_mtu = ifr->ifr_mtu;
691 struct sockaddr **llsa,
694 struct sockaddr_dl *sdl;
696 struct sockaddr_in *sin;
699 struct sockaddr_in6 *sin6;
703 switch(sa->sa_family) {
706 * No mapping needed. Just check that it's a valid MC address.
708 sdl = (struct sockaddr_dl *)sa;
709 e_addr = LLADDR(sdl);
710 if ((e_addr[0] & 1) != 1)
711 return EADDRNOTAVAIL;
717 sin = (struct sockaddr_in *)sa;
718 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
719 return EADDRNOTAVAIL;
720 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
721 sdl->sdl_len = sizeof *sdl;
722 sdl->sdl_family = AF_LINK;
723 sdl->sdl_index = ifp->if_index;
724 sdl->sdl_type = IFT_ETHER;
725 sdl->sdl_alen = ETHER_ADDR_LEN;
726 e_addr = LLADDR(sdl);
727 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
728 *llsa = (struct sockaddr *)sdl;
733 sin6 = (struct sockaddr_in6 *)sa;
734 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
736 * An IP6 address of 0 means listen to all
737 * of the Ethernet multicast address used for IP6.
738 * (This is used for multicast routers.)
740 ifp->if_flags |= IFF_ALLMULTI;
744 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
745 return EADDRNOTAVAIL;
746 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
747 sdl->sdl_len = sizeof *sdl;
748 sdl->sdl_family = AF_LINK;
749 sdl->sdl_index = ifp->if_index;
750 sdl->sdl_type = IFT_ETHER;
751 sdl->sdl_alen = ETHER_ADDR_LEN;
752 e_addr = LLADDR(sdl);
753 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
754 *llsa = (struct sockaddr *)sdl;
760 * Well, the text isn't quite right, but it's the name
769 * This is for reference. We have a table-driven version
770 * of the little-endian crc32 generator, which is faster
771 * than the double-loop.
774 ether_crc32_le(const uint8_t *buf, size_t len)
776 uint32_t c, crc, carry;
779 crc = 0xffffffffU; /* initial value */
781 for (i = 0; i < len; i++) {
783 for (j = 0; j < 8; j++) {
784 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
788 crc = (crc ^ ETHER_CRC_POLY_LE);
796 ether_crc32_le(const uint8_t *buf, size_t len)
798 static const uint32_t crctab[] = {
799 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
800 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
801 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
802 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
807 crc = 0xffffffffU; /* initial value */
809 for (i = 0; i < len; i++) {
811 crc = (crc >> 4) ^ crctab[crc & 0xf];
812 crc = (crc >> 4) ^ crctab[crc & 0xf];
820 ether_crc32_be(const uint8_t *buf, size_t len)
822 uint32_t c, crc, carry;
825 crc = 0xffffffffU; /* initial value */
827 for (i = 0; i < len; i++) {
829 for (j = 0; j < 8; j++) {
830 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
834 crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
842 * find the size of ethernet header, and call classifier
845 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
846 struct altq_pktattr *pktattr)
848 struct ether_header *eh;
850 int hlen, af, hdrsize;
853 hlen = sizeof(struct ether_header);
854 eh = mtod(m, struct ether_header *);
856 ether_type = ntohs(eh->ether_type);
857 if (ether_type < ETHERMTU) {
859 struct llc *llc = (struct llc *)(eh + 1);
862 if (m->m_len < hlen ||
863 llc->llc_dsap != LLC_SNAP_LSAP ||
864 llc->llc_ssap != LLC_SNAP_LSAP ||
865 llc->llc_control != LLC_UI)
866 goto bad; /* not snap! */
868 ether_type = ntohs(llc->llc_un.type_snap.ether_type);
871 if (ether_type == ETHERTYPE_IP) {
873 hdrsize = 20; /* sizeof(struct ip) */
875 } else if (ether_type == ETHERTYPE_IPV6) {
877 hdrsize = 40; /* sizeof(struct ip6_hdr) */
882 while (m->m_len <= hlen) {
886 hdr = m->m_data + hlen;
887 if (m->m_len < hlen + hdrsize) {
889 * ip header is not in a single mbuf. this should not
890 * happen in the current code.
891 * (todo: use m_pulldown in the future)
897 ifq_classify(ifq, m, af, pktattr);
904 pktattr->pattr_class = NULL;
905 pktattr->pattr_hdr = NULL;
906 pktattr->pattr_af = AF_UNSPEC;
910 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
911 const struct ether_header *save_eh)
913 struct mbuf *m = *m0;
918 * Prepend the header, optimize for the common case of
919 * eh pointing into the mbuf.
921 if ((const void *)(eh + 1) == (void *)m->m_data) {
922 m->m_data -= ETHER_HDR_LEN;
923 m->m_len += ETHER_HDR_LEN;
924 m->m_pkthdr.len += ETHER_HDR_LEN;
928 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
930 bcopy(save_eh, mtod(m, struct ether_header *),
938 * Upper layer processing for a received Ethernet packet.
941 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
943 struct ether_header *eh;
944 int isr, discard = 0;
946 struct ip_fw *rule = NULL;
949 KASSERT(m->m_len >= ETHER_HDR_LEN,
950 ("ether header is not contiguous!"));
952 eh = mtod(m, struct ether_header *);
954 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
957 /* Extract info from dummynet tag */
958 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
959 KKASSERT(mtag != NULL);
960 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
961 KKASSERT(rule != NULL);
963 m_tag_delete(m, mtag);
964 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
966 /* packet is passing the second time */
971 * We got a packet which was unicast to a different Ethernet
972 * address. If the driver is working properly, then this
973 * situation can only happen when the interface is in
974 * promiscuous mode. We defer the packet discarding until the
975 * vlan processing is done, so that vlan/bridge or vlan/netgraph
978 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
979 !ETHER_IS_MULTICAST(eh->ether_dhost) &&
980 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
981 if (ether_debug & 1) {
982 kprintf("%02x:%02x:%02x:%02x:%02x:%02x "
983 "%02x:%02x:%02x:%02x:%02x:%02x "
984 "%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n",
998 ((u_char *)IFP2AC(ifp)->ac_enaddr)[0],
999 ((u_char *)IFP2AC(ifp)->ac_enaddr)[1],
1000 ((u_char *)IFP2AC(ifp)->ac_enaddr)[2],
1001 ((u_char *)IFP2AC(ifp)->ac_enaddr)[3],
1002 ((u_char *)IFP2AC(ifp)->ac_enaddr)[4],
1003 ((u_char *)IFP2AC(ifp)->ac_enaddr)[5]
1006 if ((ether_debug & 2) == 0)
1011 if (IPFW_LOADED && ether_ipfw != 0 && !discard) {
1012 struct ether_header save_eh = *eh;
1014 /* XXX old crufty stuff, needs to be removed */
1015 m_adj(m, sizeof(struct ether_header));
1017 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
1022 ether_restore_header(&m, eh, &save_eh);
1025 eh = mtod(m, struct ether_header *);
1028 ether_type = ntohs(eh->ether_type);
1029 KKASSERT(ether_type != ETHERTYPE_VLAN);
1031 if (m->m_flags & M_VLANTAG) {
1032 void (*vlan_input_func)(struct mbuf *);
1034 vlan_input_func = vlan_input_p;
1035 if (vlan_input_func != NULL) {
1038 m->m_pkthdr.rcvif->if_noproto++;
1045 * If we have been asked to discard this packet
1046 * (e.g. not for us), drop it before entering
1055 * Clear protocol specific flags,
1056 * before entering the upper layer.
1058 m->m_flags &= ~M_ETHER_FLAGS;
1060 /* Strip ethernet header. */
1061 m_adj(m, sizeof(struct ether_header));
1063 switch (ether_type) {
1066 if ((m->m_flags & M_LENCHECKED) == 0) {
1067 if (!ip_lengthcheck(&m, 0))
1070 if (ipflow_fastforward(m))
1076 if (ifp->if_flags & IFF_NOARP) {
1077 /* Discard packet if ARP is disabled on interface */
1086 case ETHERTYPE_IPV6:
1095 * Hold BGL and recheck ef_inputp
1098 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) {
1109 case ETHERTYPE_MPLS:
1110 case ETHERTYPE_MPLS_MCAST:
1111 /* Should have been set by ether_input_pkt(). */
1112 KKASSERT(m->m_flags & M_MPLSLABELED);
1119 * The accurate msgport is not determined before
1120 * we reach here, so recharacterize packet.
1122 m->m_flags &= ~M_HASH;
1126 * Hold BGL and recheck ef_inputp
1129 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) {
1136 if (ng_ether_input_orphan_p != NULL) {
1138 * Put back the ethernet header so netgraph has a
1139 * consistent view of inbound packets.
1141 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
1144 * M_PREPEND frees the mbuf in case of failure.
1149 * Hold BGL and recheck ng_ether_input_orphan_p
1152 if (ng_ether_input_orphan_p != NULL) {
1153 ng_ether_input_orphan_p(ifp, m);
1163 if (m->m_flags & M_HASH) {
1164 if (&curthread->td_msgport == cpu_portfn(m->m_pkthdr.hash)) {
1165 netisr_handle(isr, m);
1169 * XXX Something is wrong,
1170 * we probably should panic here!
1172 m->m_flags &= ~M_HASH;
1173 atomic_add_long(ðer_input_wronghash, 1);
1177 atomic_add_long(ðer_input_requeue, 1);
1179 netisr_queue(isr, m);
1183 * First we perform any link layer operations, then continue to the
1184 * upper layers with ether_demux_oncpu().
1187 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1193 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1195 * Receiving interface's flags are changed, when this
1196 * packet is waiting for processing; discard it.
1203 * Tap the packet off here for a bridge. bridge_input()
1204 * will return NULL if it has consumed the packet, otherwise
1205 * it gets processed as normal. Note that bridge_input()
1206 * will always return the original packet if we need to
1207 * process it locally.
1209 if (ifp->if_bridge) {
1210 KASSERT(bridge_input_p != NULL,
1211 ("%s: if_bridge not loaded!", __func__));
1213 if(m->m_flags & M_ETHER_BRIDGED) {
1214 m->m_flags &= ~M_ETHER_BRIDGED;
1216 m = bridge_input_p(ifp, m);
1220 KASSERT(ifp == m->m_pkthdr.rcvif,
1221 ("bridge_input_p changed rcvif"));
1226 carp = ifp->if_carp;
1228 m = carp_input(carp, m);
1231 KASSERT(ifp == m->m_pkthdr.rcvif,
1232 ("carp_input changed rcvif"));
1236 /* Handle ng_ether(4) processing, if any */
1237 if (ng_ether_input_p != NULL) {
1239 * Hold BGL and recheck ng_ether_input_p
1242 if (ng_ether_input_p != NULL)
1243 ng_ether_input_p(ifp, &m);
1250 /* Continue with upper layer processing */
1251 ether_demux_oncpu(ifp, m);
1255 * Perform certain functions of ether_input_pkt():
1257 * - Update statistics
1258 * - Run bpf(4) tap if requested
1259 * Then pass the packet to ether_input_oncpu().
1261 * This function should be used by pseudo interface (e.g. vlan(4)),
1262 * when it tries to claim that the packet is received by it.
1268 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags)
1270 /* Discard packet if interface is not up */
1271 if (!(ifp->if_flags & IFF_UP)) {
1277 * Change receiving interface. The bridge will often pass a flag to
1278 * ask that this not be done so ARPs get applied to the correct
1281 if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 ||
1282 m->m_pkthdr.rcvif == NULL) {
1283 m->m_pkthdr.rcvif = ifp;
1286 /* Update statistics */
1288 ifp->if_ibytes += m->m_pkthdr.len;
1289 if (m->m_flags & (M_MCAST | M_BCAST))
1292 if (reinput_flags & REINPUT_RUNBPF)
1295 ether_input_oncpu(ifp, m);
1298 static __inline boolean_t
1299 ether_vlancheck(struct mbuf **m0)
1301 struct mbuf *m = *m0;
1302 struct ether_header *eh;
1303 uint16_t ether_type;
1305 eh = mtod(m, struct ether_header *);
1306 ether_type = ntohs(eh->ether_type);
1308 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) {
1310 * Extract vlan tag if hardware does not do it for us
1312 vlan_ether_decap(&m);
1316 eh = mtod(m, struct ether_header *);
1317 ether_type = ntohs(eh->ether_type);
1320 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) {
1322 * To prevent possible dangerous recursion,
1323 * we don't do vlan-in-vlan
1325 m->m_pkthdr.rcvif->if_noproto++;
1328 KKASSERT(ether_type != ETHERTYPE_VLAN);
1330 m->m_flags |= M_ETHER_VLANCHECKED;
1341 ether_input_handler(netmsg_t nmsg)
1343 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */
1344 struct ether_header *eh;
1350 ifp = m->m_pkthdr.rcvif;
1352 eh = mtod(m, struct ether_header *);
1353 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1354 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1355 ifp->if_addrlen) == 0)
1356 m->m_flags |= M_BCAST;
1358 m->m_flags |= M_MCAST;
1362 if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) {
1363 if (!ether_vlancheck(&m)) {
1364 KKASSERT(m == NULL);
1369 ether_input_oncpu(ifp, m);
1373 * Send the packet to the target msgport
1375 * At this point the packet had better be characterized (M_HASH set),
1376 * so we know which cpu to send it to.
1379 ether_dispatch(int isr, struct mbuf *m)
1381 struct netmsg_packet *pmsg;
1383 KKASSERT(m->m_flags & M_HASH);
1384 pmsg = &m->m_hdr.mh_netmsg;
1385 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1386 0, ether_input_handler);
1387 pmsg->nm_packet = m;
1388 pmsg->base.lmsg.u.ms_result = isr;
1390 logether(disp_beg, NULL);
1391 lwkt_sendmsg(cpu_portfn(m->m_pkthdr.hash), &pmsg->base.lmsg);
1392 logether(disp_end, NULL);
1396 * Process a received Ethernet packet.
1398 * The ethernet header is assumed to be in the mbuf so the caller
1399 * MUST MAKE SURE that there are at least sizeof(struct ether_header)
1400 * bytes in the first mbuf.
1403 ether_input_pkt(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi)
1409 /* Discard packet if interface is not up */
1410 if (!(ifp->if_flags & IFF_UP)) {
1415 if (m->m_len < sizeof(struct ether_header)) {
1416 /* XXX error in the caller. */
1421 m->m_pkthdr.rcvif = ifp;
1423 logether(pkt_beg, ifp);
1425 ETHER_BPF_MTAP(ifp, m);
1427 ifp->if_ibytes += m->m_pkthdr.len;
1429 if (ifp->if_flags & IFF_MONITOR) {
1430 struct ether_header *eh;
1432 eh = mtod(m, struct ether_header *);
1433 if (ETHER_IS_MULTICAST(eh->ether_dhost))
1437 * Interface marked for monitoring; discard packet.
1441 logether(pkt_end, ifp);
1446 * If the packet has been characterized (pi->pi_netisr / M_HASH)
1447 * we can dispatch it immediately without further inspection.
1449 if (pi != NULL && (m->m_flags & M_HASH)) {
1451 atomic_add_long(ðer_pktinfo_try, 1);
1453 netisr_hashcheck(pi->pi_netisr, m, pi);
1454 if (m->m_flags & M_HASH) {
1455 ether_dispatch(pi->pi_netisr, m);
1457 atomic_add_long(ðer_pktinfo_hit, 1);
1459 logether(pkt_end, ifp);
1464 else if (ifp->if_capenable & IFCAP_RSS) {
1466 atomic_add_long(ðer_rss_nopi, 1);
1468 atomic_add_long(ðer_rss_nohash, 1);
1473 * Packet hash will be recalculated by software,
1474 * so clear the M_HASH flag set by the driver;
1475 * the hash value calculated by the hardware may
1476 * not be exactly what we want.
1478 m->m_flags &= ~M_HASH;
1480 if (!ether_vlancheck(&m)) {
1481 KKASSERT(m == NULL);
1482 logether(pkt_end, ifp);
1486 isr = ether_characterize(&m);
1488 logether(pkt_end, ifp);
1493 * Finally dispatch it
1495 ether_dispatch(isr, m);
1497 logether(pkt_end, ifp);
1501 ether_characterize(struct mbuf **m0)
1503 struct mbuf *m = *m0;
1504 struct ether_header *eh;
1505 uint16_t ether_type;
1508 eh = mtod(m, struct ether_header *);
1509 ether_type = ntohs(eh->ether_type);
1512 * Map ether type to netisr id.
1514 switch (ether_type) {
1526 case ETHERTYPE_IPV6:
1538 case ETHERTYPE_MPLS:
1539 case ETHERTYPE_MPLS_MCAST:
1540 m->m_flags |= M_MPLSLABELED;
1547 * NETISR_MAX is an invalid value; it is chosen to let
1548 * netisr_characterize() know that we have no clear
1549 * idea where this packet should go.
1556 * Ask the isr to characterize the packet since we couldn't.
1557 * This is an attempt to optimally get us onto the correct protocol
1560 netisr_characterize(isr, &m, sizeof(struct ether_header));
1567 ether_demux_handler(netmsg_t nmsg)
1569 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */
1575 ifp = m->m_pkthdr.rcvif;
1577 ether_demux_oncpu(ifp, m);
1581 ether_demux(struct mbuf *m)
1583 struct netmsg_packet *pmsg;
1586 isr = ether_characterize(&m);
1590 KKASSERT(m->m_flags & M_HASH);
1591 pmsg = &m->m_hdr.mh_netmsg;
1592 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1593 0, ether_demux_handler);
1594 pmsg->nm_packet = m;
1595 pmsg->base.lmsg.u.ms_result = isr;
1597 lwkt_sendmsg(cpu_portfn(m->m_pkthdr.hash), &pmsg->base.lmsg);
1601 ether_tso_pullup(struct mbuf **mp, int *hoff0, struct ip **ip, int *iphlen,
1602 struct tcphdr **th, int *thoff)
1604 struct mbuf *m = *mp;
1605 struct ether_header *eh;
1609 KASSERT(M_WRITABLE(m), ("not writable"));
1611 hoff = ETHER_HDR_LEN;
1612 if (m->m_len < hoff) {
1613 m = m_pullup(m, hoff);
1617 eh = mtod(m, struct ether_header *);
1618 type = eh->ether_type;
1620 if (type == htons(ETHERTYPE_VLAN)) {
1621 struct ether_vlan_header *evh;
1623 hoff += EVL_ENCAPLEN;
1624 if (m->m_len < hoff) {
1625 m = m_pullup(m, hoff);
1629 evh = mtod(m, struct ether_vlan_header *);
1630 type = evh->evl_proto;
1632 KASSERT(type == htons(ETHERTYPE_IP), ("not IP %d", ntohs(type)));
1636 return tcp_tso_pullup(mp, hoff, ip, iphlen, th, thoff);
1645 MODULE_VERSION(ether, 1);