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. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)if_ethersubr.c 8.1 (Berkeley) 6/10/93
30 * $FreeBSD: src/sys/net/if_ethersubr.c,v 1.70.2.33 2003/04/28 15:45:53 archie Exp $
34 #include "opt_inet6.h"
36 #include "opt_netgraph.h"
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/globaldata.h>
43 #include <sys/kernel.h>
46 #include <sys/malloc.h>
48 #include <sys/msgport.h>
49 #include <sys/socket.h>
50 #include <sys/sockio.h>
51 #include <sys/sysctl.h>
52 #include <sys/thread.h>
54 #include <sys/thread2.h>
55 #include <sys/mplock2.h>
58 #include <net/netisr.h>
59 #include <net/route.h>
60 #include <net/if_llc.h>
61 #include <net/if_dl.h>
62 #include <net/if_types.h>
63 #include <net/ifq_var.h>
65 #include <net/ethernet.h>
66 #include <net/vlan/if_vlan_ether.h>
67 #include <net/vlan/if_vlan_var.h>
68 #include <net/netmsg2.h>
69 #include <net/netisr2.h>
71 #if defined(INET) || defined(INET6)
72 #include <netinet/in.h>
73 #include <netinet/ip_var.h>
74 #include <netinet/tcp_var.h>
75 #include <netinet/if_ether.h>
76 #include <netinet/ip_flow.h>
77 #include <net/ipfw/ip_fw.h>
78 #include <net/ipfw3/ip_fw.h>
79 #include <net/dummynet/ip_dummynet.h>
82 #include <netinet6/nd6.h>
86 #include <netinet/ip_carp.h>
90 #include <netproto/mpls/mpls.h>
93 /* netgraph node hooks for ng_ether(4) */
94 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
95 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
96 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
97 void (*ng_ether_attach_p)(struct ifnet *ifp);
98 void (*ng_ether_detach_p)(struct ifnet *ifp);
100 void (*vlan_input_p)(struct mbuf *);
102 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
104 static void ether_restore_header(struct mbuf **, const struct ether_header *,
105 const struct ether_header *);
106 static int ether_characterize(struct mbuf **);
107 static void ether_dispatch(struct ifnet *, int, struct mbuf *, int);
112 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
113 int (*bridge_output_p)(struct ifnet *, struct mbuf *);
114 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
115 struct ifnet *(*bridge_interface_p)(void *if_bridge);
117 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
123 void (*lagg_input_p)(struct ifnet *, struct mbuf *);
124 int (*lagg_output_p)(struct ifnet *, struct mbuf *);
126 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
127 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
130 #define gotoerr(e) do { error = (e); goto bad; } while (0)
131 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
133 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
135 const struct ether_header *eh);
137 static int ether_ipfw;
138 static u_long ether_restore_hdr;
139 static u_long ether_prepend_hdr;
140 static u_long ether_input_wronghash;
141 static int ether_debug;
144 static u_long ether_pktinfo_try;
145 static u_long ether_pktinfo_hit;
146 static u_long ether_rss_nopi;
147 static u_long ether_rss_nohash;
148 static u_long ether_input_requeue;
150 static u_long ether_input_wronghwhash;
151 static int ether_input_ckhash;
153 #define ETHER_TSOLEN_DEFAULT (4 * ETHERMTU)
155 #define ETHER_NMBCLUSTERS_DEFMIN 32
156 #define ETHER_NMBCLUSTERS_DEFAULT 256
158 static int ether_tsolen_default = ETHER_TSOLEN_DEFAULT;
159 TUNABLE_INT("net.link.ether.tsolen", ðer_tsolen_default);
161 static int ether_nmbclusters_default = ETHER_NMBCLUSTERS_DEFAULT;
162 TUNABLE_INT("net.link.ether.nmbclusters", ðer_nmbclusters_default);
164 SYSCTL_DECL(_net_link);
165 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
166 SYSCTL_INT(_net_link_ether, OID_AUTO, debug, CTLFLAG_RW,
167 ðer_debug, 0, "Ether debug");
168 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
169 ðer_ipfw, 0, "Pass ether pkts through firewall");
170 SYSCTL_ULONG(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW,
171 ðer_restore_hdr, 0, "# of ether header restoration");
172 SYSCTL_ULONG(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW,
173 ðer_prepend_hdr, 0,
174 "# of ether header restoration which prepends mbuf");
175 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghash, CTLFLAG_RW,
176 ðer_input_wronghash, 0, "# of input packets with wrong hash");
177 SYSCTL_INT(_net_link_ether, OID_AUTO, tsolen, CTLFLAG_RW,
178 ðer_tsolen_default, 0, "Default max TSO length");
181 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nopi, CTLFLAG_RW,
182 ðer_rss_nopi, 0, "# of packets do not have pktinfo");
183 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nohash, CTLFLAG_RW,
184 ðer_rss_nohash, 0, "# of packets do not have hash");
185 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_try, CTLFLAG_RW,
186 ðer_pktinfo_try, 0,
187 "# of tries to find packets' msgport using pktinfo");
188 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_hit, CTLFLAG_RW,
189 ðer_pktinfo_hit, 0,
190 "# of packets whose msgport are found using pktinfo");
191 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_requeue, CTLFLAG_RW,
192 ðer_input_requeue, 0, "# of input packets gets requeued");
194 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghwhash, CTLFLAG_RW,
195 ðer_input_wronghwhash, 0, "# of input packets with wrong hw hash");
196 SYSCTL_INT(_net_link_ether, OID_AUTO, always_ckhash, CTLFLAG_RW,
197 ðer_input_ckhash, 0, "always check hash");
199 #define ETHER_KTR_STR "ifp=%p"
200 #define ETHER_KTR_ARGS struct ifnet *ifp
202 #define KTR_ETHERNET KTR_ALL
204 KTR_INFO_MASTER(ether);
205 KTR_INFO(KTR_ETHERNET, ether, pkt_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARGS);
206 KTR_INFO(KTR_ETHERNET, ether, pkt_end, 1, ETHER_KTR_STR, ETHER_KTR_ARGS);
207 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARGS);
208 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARGS);
209 #define logether(name, arg) KTR_LOG(ether_ ## name, arg)
212 * Ethernet output routine.
213 * Encapsulate a packet of type family for the local net.
214 * Use trailer local net encapsulation if enough data in first
215 * packet leaves a multiple of 512 bytes of data in remainder.
216 * Assumes that ifp is actually pointer to arpcom structure.
219 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
222 struct ether_header *eh, *deh;
225 int hlen = ETHER_HDR_LEN; /* link layer header length */
226 struct arpcom *ac = IFP2AC(ifp);
229 ASSERT_NETISR_NCPUS(mycpuid);
230 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
232 if (ifp->if_flags & IFF_MONITOR)
234 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
237 M_PREPEND(m, sizeof(struct ether_header), M_NOWAIT);
240 m->m_pkthdr.csum_lhlen = sizeof(struct ether_header);
241 eh = mtod(m, struct ether_header *);
242 edst = eh->ether_dhost;
245 * Fill in the destination ethernet address and frame type.
247 switch (dst->sa_family) {
250 error = arpresolve(ifp, rt, m, dst, edst);
252 return error == EWOULDBLOCK ? 0 : error;
254 if (m->m_flags & M_MPLSLABELED)
255 eh->ether_type = htons(ETHERTYPE_MPLS);
258 eh->ether_type = htons(ETHERTYPE_IP);
263 error = nd6_resolve(&ac->ac_if, rt, m, dst, edst);
265 return error == EWOULDBLOCK ? 0 : error;
266 eh->ether_type = htons(ETHERTYPE_IPV6);
269 case pseudo_AF_HDRCMPLT:
271 loop_copy = -1; /* if this is for us, don't do it */
272 deh = (struct ether_header *)dst->sa_data;
273 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
274 eh->ether_type = deh->ether_type;
278 if_printf(ifp, "can't handle af%d\n", dst->sa_family);
279 gotoerr(EAFNOSUPPORT);
282 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */
283 memcpy(eh->ether_shost,
284 ((struct ether_header *)dst->sa_data)->ether_shost,
287 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
290 * Bridges require special output handling.
292 if (ifp->if_bridge) {
293 KASSERT(bridge_output_p != NULL,
294 ("%s: if_bridge not loaded!", __func__));
295 return bridge_output_p(ifp, m);
299 KASSERT(lagg_output_p != NULL,
300 ("%s: if_lagg not loaded!", __func__));
301 return lagg_output_p(ifp, m);
306 * If a simplex interface, and the packet is being sent to our
307 * Ethernet address or a broadcast address, loopback a copy.
308 * XXX To make a simplex device behave exactly like a duplex
309 * device, we should copy in the case of sending to our own
310 * ethernet address (thus letting the original actually appear
311 * on the wire). However, we don't do that here for security
312 * reasons and compatibility with the original behavior.
314 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
317 if (m->m_pkthdr.csum_flags & CSUM_IP)
318 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
319 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
320 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
321 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
324 if ((n = m_copypacket(m, M_NOWAIT)) != NULL) {
325 n->m_pkthdr.csum_flags |= csum_flags;
326 if (csum_flags & CSUM_DATA_VALID)
327 n->m_pkthdr.csum_data = 0xffff;
328 if_simloop(ifp, n, dst->sa_family, hlen);
330 IFNET_STAT_INC(ifp, iqdrops, 1);
331 } else if (bcmp(eh->ether_dhost, eh->ether_shost,
332 ETHER_ADDR_LEN) == 0) {
333 m->m_pkthdr.csum_flags |= csum_flags;
334 if (csum_flags & CSUM_DATA_VALID)
335 m->m_pkthdr.csum_data = 0xffff;
336 if_simloop(ifp, m, dst->sa_family, hlen);
337 return (0); /* XXX */
342 if (ifp->if_type == IFT_CARP) {
343 ifp = carp_parent(ifp);
345 gotoerr(ENETUNREACH);
350 * Check precondition again
352 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
354 if (ifp->if_flags & IFF_MONITOR)
356 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
357 (IFF_UP | IFF_RUNNING))
362 /* Handle ng_ether(4) processing, if any */
363 if (ng_ether_output_p != NULL) {
365 * Hold BGL and recheck ng_ether_output_p
368 if (ng_ether_output_p != NULL) {
369 if ((error = ng_ether_output_p(ifp, &m)) != 0) {
381 /* Continue with link-layer output */
382 return ether_output_frame(ifp, m);
390 * Returns the bridge interface an ifp is associated
393 * Only call if ifp->if_bridge != NULL.
396 ether_bridge_interface(struct ifnet *ifp)
398 if (bridge_interface_p)
399 return(bridge_interface_p(ifp->if_bridge));
404 * Ethernet link layer output routine to send a raw frame to the device.
406 * This assumes that the 14 byte Ethernet header is present and contiguous
410 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
412 struct ip_fw *rule = NULL;
414 struct altq_pktattr pktattr;
416 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
418 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
421 /* Extract info from dummynet tag */
422 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
423 KKASSERT(mtag != NULL);
424 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
425 KKASSERT(rule != NULL);
427 m_tag_delete(m, mtag);
428 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
431 if (ifq_is_enabled(&ifp->if_snd))
432 altq_etherclassify(&ifp->if_snd, m, &pktattr);
434 if ((IPFW_LOADED || IPFW3_LOADED) && ether_ipfw != 0) {
435 struct ether_header save_eh, *eh;
437 eh = mtod(m, struct ether_header *);
439 m_adj(m, ETHER_HDR_LEN);
440 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) {
444 return ENOBUFS; /* pkt dropped */
446 return 0; /* consumed e.g. in a pipe */
449 /* packet was ok, restore the ethernet header */
450 ether_restore_header(&m, eh, &save_eh);
459 * Queue message on interface, update output statistics if
460 * successful, and start output if interface not yet active.
462 error = ifq_dispatch(ifp, m, &pktattr);
467 * ipfw processing for ethernet packets (in and out).
468 * The second parameter is NULL from ether_demux(), and ifp from
469 * ether_output_frame().
472 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule,
473 const struct ether_header *eh)
475 struct ether_header save_eh = *eh; /* might be a ptr in *m0 */
476 struct ip_fw_args args;
481 if (*rule != NULL && fw_one_pass)
482 return TRUE; /* dummynet packet, already partially processed */
485 * I need some amount of data to be contiguous.
487 i = min((*m0)->m_pkthdr.len, max_protohdr);
488 if ((*m0)->m_len < i) {
489 *m0 = m_pullup(*m0, i);
497 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
498 m_tag_delete(*m0, mtag);
499 if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
500 mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL);
501 KKASSERT(mtag != NULL);
502 m_tag_delete(*m0, mtag);
503 (*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED;
508 args.m = *m0; /* the packet we are looking at */
509 args.oif = dst; /* destination, if any */
510 args.rule = *rule; /* matching rule to restart */
511 args.eh = &save_eh; /* MAC header for bridged/MAC packets */
512 i = ip_fw_chk_ptr(&args);
527 * XXX at some point add support for divert/forward actions.
528 * If none of the above matches, we have to drop the pkt.
534 * Pass the pkt to dummynet, which consumes it.
536 m = *m0; /* pass the original to dummynet */
537 *m0 = NULL; /* and nothing back to the caller */
539 ether_restore_header(&m, eh, &save_eh);
543 m = ip_fw_dn_io_ptr(m, args.cookie,
544 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
550 panic("unknown ipfw return value: %d", i);
555 * Perform common duties while attaching to interface list
558 ether_ifattach(struct ifnet *ifp, const uint8_t *lla,
559 lwkt_serialize_t serializer)
561 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
566 ether_ifattach_bpf(struct ifnet *ifp, const uint8_t *lla,
567 u_int dlt, u_int hdrlen, lwkt_serialize_t serializer)
569 struct sockaddr_dl *sdl;
570 char ethstr[ETHER_ADDRSTRLEN + 1];
575 * If driver does not configure # of mbuf clusters/jclusters
576 * that could sit on the device queues for quite some time,
578 * - The device queues only consume mbuf clusters.
579 * - No more than ether_nmbclusters_default (by default 256)
580 * mbuf clusters will sit on the device queues for quite
583 if (ifp->if_nmbclusters <= 0 && ifp->if_nmbjclusters <= 0) {
584 if (ether_nmbclusters_default < ETHER_NMBCLUSTERS_DEFMIN) {
585 kprintf("ether nmbclusters %d -> %d\n",
586 ether_nmbclusters_default,
587 ETHER_NMBCLUSTERS_DEFAULT);
588 ether_nmbclusters_default = ETHER_NMBCLUSTERS_DEFAULT;
590 ifp->if_nmbclusters = ether_nmbclusters_default;
593 ifp->if_type = IFT_ETHER;
594 ifp->if_addrlen = ETHER_ADDR_LEN;
595 ifp->if_hdrlen = ETHER_HDR_LEN;
596 if_attach(ifp, serializer);
598 for (i = 0; i < ifq->altq_subq_cnt; ++i) {
599 struct ifaltq_subque *ifsq = ifq_get_subq(ifq, i);
601 ifsq->ifsq_maxbcnt = ifsq->ifsq_maxlen *
602 (ETHER_MAX_LEN - ETHER_CRC_LEN);
604 ifp->if_mtu = ETHERMTU;
605 if (ifp->if_tsolen <= 0) {
606 if ((ether_tsolen_default / ETHERMTU) < 2) {
607 kprintf("ether TSO maxlen %d -> %d\n",
608 ether_tsolen_default, ETHER_TSOLEN_DEFAULT);
609 ether_tsolen_default = ETHER_TSOLEN_DEFAULT;
611 ifp->if_tsolen = ether_tsolen_default;
613 if (ifp->if_baudrate == 0)
614 ifp->if_baudrate = 10000000;
615 ifp->if_output = ether_output;
616 ifp->if_input = ether_input;
617 ifp->if_resolvemulti = ether_resolvemulti;
618 ifp->if_broadcastaddr = etherbroadcastaddr;
619 sdl = IF_LLSOCKADDR(ifp);
620 sdl->sdl_type = IFT_ETHER;
621 sdl->sdl_alen = ifp->if_addrlen;
622 bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
624 * XXX Keep the current drivers happy.
625 * XXX Remove once all drivers have been cleaned up
627 if (lla != IFP2AC(ifp)->ac_enaddr)
628 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
629 bpfattach(ifp, dlt, hdrlen);
630 if (ng_ether_attach_p != NULL)
631 (*ng_ether_attach_p)(ifp);
633 if_printf(ifp, "MAC address: %s\n", kether_ntoa(lla, ethstr));
637 * Perform common duties while detaching an Ethernet interface
640 ether_ifdetach(struct ifnet *ifp)
644 if (ng_ether_detach_p != NULL)
645 (*ng_ether_detach_p)(ifp);
651 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
653 struct ifaddr *ifa = (struct ifaddr *) data;
654 struct ifreq *ifr = (struct ifreq *) data;
657 #define IF_INIT(ifp) \
659 if (((ifp)->if_flags & IFF_UP) == 0) { \
660 (ifp)->if_flags |= IFF_UP; \
661 (ifp)->if_init((ifp)->if_softc); \
665 ASSERT_IFNET_SERIALIZED_ALL(ifp);
669 switch (ifa->ifa_addr->sa_family) {
672 IF_INIT(ifp); /* before arpwhohas */
673 arp_ifinit(ifp, ifa);
683 bcopy(IFP2AC(ifp)->ac_enaddr,
684 ((struct sockaddr *)ifr->ifr_data)->sa_data,
690 * Set the interface MTU.
692 if (ifr->ifr_mtu > ETHERMTU) {
695 ifp->if_mtu = ifr->ifr_mtu;
710 struct sockaddr **llsa,
713 struct sockaddr_dl *sdl;
715 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 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
740 sdl->sdl_len = sizeof *sdl;
741 sdl->sdl_family = AF_LINK;
742 sdl->sdl_index = ifp->if_index;
743 sdl->sdl_type = IFT_ETHER;
744 sdl->sdl_alen = ETHER_ADDR_LEN;
745 e_addr = LLADDR(sdl);
746 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
747 *llsa = (struct sockaddr *)sdl;
752 sin6 = (struct sockaddr_in6 *)sa;
753 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
755 * An IP6 address of 0 means listen to all
756 * of the Ethernet multicast address used for IP6.
757 * (This is used for multicast routers.)
759 ifp->if_flags |= IFF_ALLMULTI;
763 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
764 return EADDRNOTAVAIL;
765 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
766 sdl->sdl_len = sizeof *sdl;
767 sdl->sdl_family = AF_LINK;
768 sdl->sdl_index = ifp->if_index;
769 sdl->sdl_type = IFT_ETHER;
770 sdl->sdl_alen = ETHER_ADDR_LEN;
771 e_addr = LLADDR(sdl);
772 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
773 *llsa = (struct sockaddr *)sdl;
779 * Well, the text isn't quite right, but it's the name
788 * This is for reference. We have a table-driven version
789 * of the little-endian crc32 generator, which is faster
790 * than the double-loop.
793 ether_crc32_le(const uint8_t *buf, size_t len)
795 uint32_t c, crc, carry;
798 crc = 0xffffffffU; /* initial value */
800 for (i = 0; i < len; i++) {
802 for (j = 0; j < 8; j++) {
803 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
807 crc = (crc ^ ETHER_CRC_POLY_LE);
815 ether_crc32_le(const uint8_t *buf, size_t len)
817 static const uint32_t crctab[] = {
818 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
819 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
820 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
821 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
826 crc = 0xffffffffU; /* initial value */
828 for (i = 0; i < len; i++) {
830 crc = (crc >> 4) ^ crctab[crc & 0xf];
831 crc = (crc >> 4) ^ crctab[crc & 0xf];
839 ether_crc32_be(const uint8_t *buf, size_t len)
841 uint32_t c, crc, carry;
844 crc = 0xffffffffU; /* initial value */
846 for (i = 0; i < len; i++) {
848 for (j = 0; j < 8; j++) {
849 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
853 crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
861 * find the size of ethernet header, and call classifier
864 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
865 struct altq_pktattr *pktattr)
867 struct ether_header *eh;
869 int hlen, af, hdrsize;
871 hlen = sizeof(struct ether_header);
872 eh = mtod(m, struct ether_header *);
874 ether_type = ntohs(eh->ether_type);
875 if (ether_type < ETHERMTU) {
877 struct llc *llc = (struct llc *)(eh + 1);
880 if (m->m_len < hlen ||
881 llc->llc_dsap != LLC_SNAP_LSAP ||
882 llc->llc_ssap != LLC_SNAP_LSAP ||
883 llc->llc_control != LLC_UI)
884 goto bad; /* not snap! */
886 ether_type = ntohs(llc->llc_un.type_snap.ether_type);
889 if (ether_type == ETHERTYPE_IP) {
891 hdrsize = 20; /* sizeof(struct ip) */
893 } else if (ether_type == ETHERTYPE_IPV6) {
895 hdrsize = 40; /* sizeof(struct ip6_hdr) */
900 while (m->m_len <= hlen) {
904 if (m->m_len < hlen + hdrsize) {
906 * ip header is not in a single mbuf. this should not
907 * happen in the current code.
908 * (todo: use m_pulldown in the future)
914 ifq_classify(ifq, m, af, pktattr);
921 pktattr->pattr_class = NULL;
922 pktattr->pattr_hdr = NULL;
923 pktattr->pattr_af = AF_UNSPEC;
927 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
928 const struct ether_header *save_eh)
930 struct mbuf *m = *m0;
935 * Prepend the header, optimize for the common case of
936 * eh pointing into the mbuf.
938 if ((const void *)(eh + 1) == (void *)m->m_data) {
939 m->m_data -= ETHER_HDR_LEN;
940 m->m_len += ETHER_HDR_LEN;
941 m->m_pkthdr.len += ETHER_HDR_LEN;
945 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
947 bcopy(save_eh, mtod(m, struct ether_header *),
955 * Upper layer processing for a received Ethernet packet.
958 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
960 struct ether_header *eh;
961 int isr, discard = 0;
963 struct ip_fw *rule = NULL;
966 KASSERT(m->m_len >= ETHER_HDR_LEN,
967 ("ether header is not contiguous!"));
969 eh = mtod(m, struct ether_header *);
971 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
974 /* Extract info from dummynet tag */
975 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
976 KKASSERT(mtag != NULL);
977 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
978 KKASSERT(rule != NULL);
980 m_tag_delete(m, mtag);
981 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
983 /* packet is passing the second time */
988 * We got a packet which was unicast to a different Ethernet
989 * address. If the driver is working properly, then this
990 * situation can only happen when the interface is in
991 * promiscuous mode. We defer the packet discarding until the
992 * vlan processing is done, so that vlan/bridge or vlan/netgraph
995 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
996 !ETHER_IS_MULTICAST(eh->ether_dhost) &&
997 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
998 if (ether_debug & 1) {
999 kprintf("%02x:%02x:%02x:%02x:%02x:%02x "
1000 "%02x:%02x:%02x:%02x:%02x:%02x "
1001 "%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n",
1015 ((u_char *)IFP2AC(ifp)->ac_enaddr)[0],
1016 ((u_char *)IFP2AC(ifp)->ac_enaddr)[1],
1017 ((u_char *)IFP2AC(ifp)->ac_enaddr)[2],
1018 ((u_char *)IFP2AC(ifp)->ac_enaddr)[3],
1019 ((u_char *)IFP2AC(ifp)->ac_enaddr)[4],
1020 ((u_char *)IFP2AC(ifp)->ac_enaddr)[5]
1023 if ((ether_debug & 2) == 0)
1028 if ((IPFW_LOADED || IPFW3_LOADED) && ether_ipfw != 0 && !discard) {
1029 struct ether_header save_eh = *eh;
1031 /* XXX old crufty stuff, needs to be removed */
1032 m_adj(m, sizeof(struct ether_header));
1034 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
1039 ether_restore_header(&m, eh, &save_eh);
1042 eh = mtod(m, struct ether_header *);
1045 ether_type = ntohs(eh->ether_type);
1046 KKASSERT(ether_type != ETHERTYPE_VLAN);
1048 /* Handle input from a lagg(4) port */
1049 if (ifp->if_type == IFT_IEEE8023ADLAG) {
1050 KASSERT(lagg_input_p != NULL,
1051 ("%s: if_lagg not loaded!", __func__));
1052 (*lagg_input_p)(ifp, m);
1056 if (m->m_flags & M_VLANTAG) {
1057 void (*vlan_input_func)(struct mbuf *);
1059 vlan_input_func = vlan_input_p;
1060 /* Make sure 'vlan_input_func' is really used. */
1062 if (vlan_input_func != NULL) {
1065 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1072 * If we have been asked to discard this packet
1073 * (e.g. not for us), drop it before entering
1082 * Clear protocol specific flags,
1083 * before entering the upper layer.
1085 m->m_flags &= ~M_ETHER_FLAGS;
1087 /* Strip ethernet header. */
1088 m_adj(m, sizeof(struct ether_header));
1090 switch (ether_type) {
1093 if ((m->m_flags & M_LENCHECKED) == 0) {
1094 if (!ip_lengthcheck(&m, 0))
1097 if (ipflow_fastforward(m))
1103 if (ifp->if_flags & IFF_NOARP) {
1104 /* Discard packet if ARP is disabled on interface */
1113 case ETHERTYPE_IPV6:
1119 case ETHERTYPE_MPLS:
1120 case ETHERTYPE_MPLS_MCAST:
1121 /* Should have been set by ether_input(). */
1122 KKASSERT(m->m_flags & M_MPLSLABELED);
1129 * The accurate msgport is not determined before
1130 * we reach here, so recharacterize packet.
1132 m->m_flags &= ~M_HASH;
1133 if (ng_ether_input_orphan_p != NULL) {
1135 * Put back the ethernet header so netgraph has a
1136 * consistent view of inbound packets.
1138 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
1141 * M_PREPEND frees the mbuf in case of failure.
1146 * Hold BGL and recheck ng_ether_input_orphan_p
1149 if (ng_ether_input_orphan_p != NULL) {
1150 ng_ether_input_orphan_p(ifp, m);
1160 if (m->m_flags & M_HASH) {
1161 if (&curthread->td_msgport ==
1162 netisr_hashport(m->m_pkthdr.hash)) {
1163 netisr_handle(isr, m);
1167 * XXX Something is wrong,
1168 * we probably should panic here!
1170 m->m_flags &= ~M_HASH;
1171 atomic_add_long(ðer_input_wronghash, 1);
1175 atomic_add_long(ðer_input_requeue, 1);
1177 netisr_queue(isr, m);
1181 * First we perform any link layer operations, then continue to the
1182 * upper layers with ether_demux_oncpu().
1185 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1191 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1193 * Receiving interface's flags are changed, when this
1194 * packet is waiting for processing; discard it.
1201 * Tap the packet off here for a bridge. bridge_input()
1202 * will return NULL if it has consumed the packet, otherwise
1203 * it gets processed as normal. Note that bridge_input()
1204 * will always return the original packet if we need to
1205 * process it locally.
1207 if (ifp->if_bridge) {
1208 KASSERT(bridge_input_p != NULL,
1209 ("%s: if_bridge not loaded!", __func__));
1211 if(m->m_flags & M_ETHER_BRIDGED) {
1212 m->m_flags &= ~M_ETHER_BRIDGED;
1214 m = bridge_input_p(ifp, m);
1218 KASSERT(ifp == m->m_pkthdr.rcvif,
1219 ("bridge_input_p changed rcvif"));
1224 carp = ifp->if_carp;
1226 m = carp_input(carp, m);
1229 KASSERT(ifp == m->m_pkthdr.rcvif,
1230 ("carp_input changed rcvif"));
1234 /* Handle ng_ether(4) processing, if any */
1235 if (ng_ether_input_p != NULL) {
1237 * Hold BGL and recheck ng_ether_input_p
1240 if (ng_ether_input_p != NULL)
1241 ng_ether_input_p(ifp, &m);
1248 /* Continue with upper layer processing */
1249 ether_demux_oncpu(ifp, m);
1253 * Perform certain functions of ether_input():
1255 * - Update statistics
1256 * - Run bpf(4) tap if requested
1257 * Then pass the packet to ether_input_oncpu().
1259 * This function should be used by pseudo interface (e.g. vlan(4)),
1260 * when it tries to claim that the packet is received by it.
1266 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags)
1268 /* Discard packet if interface is not up */
1269 if (!(ifp->if_flags & IFF_UP)) {
1275 * Change receiving interface. The bridge will often pass a flag to
1276 * ask that this not be done so ARPs get applied to the correct
1279 if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 ||
1280 m->m_pkthdr.rcvif == NULL) {
1281 m->m_pkthdr.rcvif = ifp;
1284 /* Update statistics */
1285 IFNET_STAT_INC(ifp, ipackets, 1);
1286 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1287 if (m->m_flags & (M_MCAST | M_BCAST))
1288 IFNET_STAT_INC(ifp, imcasts, 1);
1290 if (reinput_flags & REINPUT_RUNBPF)
1293 ether_input_oncpu(ifp, m);
1296 static __inline boolean_t
1297 ether_vlancheck(struct mbuf **m0)
1299 struct mbuf *m = *m0;
1300 struct ether_header *eh = mtod(m, struct ether_header *);
1301 uint16_t ether_type = ntohs(eh->ether_type);
1303 if (ether_type == ETHERTYPE_VLAN) {
1304 if ((m->m_flags & M_VLANTAG) == 0) {
1306 * Extract vlan tag if hardware does not do
1309 vlan_ether_decap(&m);
1313 eh = mtod(m, struct ether_header *);
1314 ether_type = ntohs(eh->ether_type);
1315 if (ether_type == ETHERTYPE_VLAN) {
1317 * To prevent possible dangerous recursion,
1318 * we don't do vlan-in-vlan.
1320 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1325 * To prevent possible dangerous recursion,
1326 * we don't do vlan-in-vlan.
1328 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1331 KKASSERT(ether_type != ETHERTYPE_VLAN);
1334 m->m_flags |= M_ETHER_VLANCHECKED;
1345 ether_input_handler(netmsg_t nmsg)
1347 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */
1348 struct ether_header *eh;
1355 if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) {
1356 if (!ether_vlancheck(&m)) {
1357 KKASSERT(m == NULL);
1362 ifp = m->m_pkthdr.rcvif;
1363 if ((m->m_flags & (M_HASH | M_CKHASH)) == (M_HASH | M_CKHASH) ||
1364 __predict_false(ether_input_ckhash)) {
1368 * Need to verify the hash supplied by the hardware
1369 * which could be wrong.
1371 m->m_flags &= ~(M_HASH | M_CKHASH);
1372 isr = ether_characterize(&m);
1375 KKASSERT(m->m_flags & M_HASH);
1377 if (netisr_hashcpu(m->m_pkthdr.hash) != mycpuid) {
1379 * Wrong hardware supplied hash; redispatch
1381 ether_dispatch(ifp, isr, m, -1);
1382 if (__predict_false(ether_input_ckhash))
1383 atomic_add_long(ðer_input_wronghwhash, 1);
1388 eh = mtod(m, struct ether_header *);
1389 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1390 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1391 ifp->if_addrlen) == 0)
1392 m->m_flags |= M_BCAST;
1394 m->m_flags |= M_MCAST;
1395 IFNET_STAT_INC(ifp, imcasts, 1);
1398 ether_input_oncpu(ifp, m);
1402 * Send the packet to the target netisr msgport
1404 * At this point the packet must be characterized (M_HASH set),
1405 * so we know which netisr to send it to.
1408 ether_dispatch(struct ifnet *ifp, int isr, struct mbuf *m, int cpuid)
1410 struct netmsg_packet *pmsg;
1413 KKASSERT(m->m_flags & M_HASH);
1414 target_cpuid = netisr_hashcpu(m->m_pkthdr.hash);
1416 pmsg = &m->m_hdr.mh_netmsg;
1417 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1418 0, ether_input_handler);
1419 pmsg->nm_packet = m;
1420 pmsg->base.lmsg.u.ms_result = isr;
1422 logether(disp_beg, NULL);
1423 if (target_cpuid == cpuid) {
1424 if ((ifp->if_flags & IFF_IDIRECT) && IN_NETISR_NCPUS(cpuid)) {
1425 ether_input_handler((netmsg_t)pmsg);
1427 lwkt_sendmsg_oncpu(netisr_cpuport(target_cpuid),
1431 lwkt_sendmsg(netisr_cpuport(target_cpuid),
1434 logether(disp_end, NULL);
1438 * Process a received Ethernet packet.
1440 * The ethernet header is assumed to be in the mbuf so the caller
1441 * MUST MAKE SURE that there are at least sizeof(struct ether_header)
1442 * bytes in the first mbuf.
1444 * If the caller knows that the current thread is stick to the current
1445 * cpu, e.g. the interrupt thread or the netisr thread, the current cpuid
1446 * (mycpuid) should be passed through 'cpuid' argument. Else -1 should
1447 * be passed as 'cpuid' argument.
1450 ether_input(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi,
1457 /* Discard packet if interface is not up */
1458 if (!(ifp->if_flags & IFF_UP)) {
1463 if (m->m_len < sizeof(struct ether_header)) {
1464 /* XXX error in the caller. */
1469 m->m_pkthdr.rcvif = ifp;
1471 logether(pkt_beg, ifp);
1473 ETHER_BPF_MTAP(ifp, m);
1475 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1477 if (ifp->if_flags & IFF_MONITOR) {
1478 struct ether_header *eh;
1480 eh = mtod(m, struct ether_header *);
1481 if (ETHER_IS_MULTICAST(eh->ether_dhost))
1482 IFNET_STAT_INC(ifp, imcasts, 1);
1485 * Interface marked for monitoring; discard packet.
1489 logether(pkt_end, ifp);
1494 * If the packet has been characterized (pi->pi_netisr / M_HASH)
1495 * we can dispatch it immediately with trivial checks.
1497 if (pi != NULL && (m->m_flags & M_HASH)) {
1499 atomic_add_long(ðer_pktinfo_try, 1);
1501 netisr_hashcheck(pi->pi_netisr, m, pi);
1502 if (m->m_flags & M_HASH) {
1503 ether_dispatch(ifp, pi->pi_netisr, m, cpuid);
1505 atomic_add_long(ðer_pktinfo_hit, 1);
1507 logether(pkt_end, ifp);
1512 else if (ifp->if_capenable & IFCAP_RSS) {
1514 atomic_add_long(ðer_rss_nopi, 1);
1516 atomic_add_long(ðer_rss_nohash, 1);
1521 * Packet hash will be recalculated by software, so clear
1522 * the M_HASH and M_CKHASH flag set by the driver; the hash
1523 * value calculated by the hardware may not be exactly what
1526 m->m_flags &= ~(M_HASH | M_CKHASH);
1528 if (!ether_vlancheck(&m)) {
1529 KKASSERT(m == NULL);
1530 logether(pkt_end, ifp);
1534 isr = ether_characterize(&m);
1536 logether(pkt_end, ifp);
1541 * Finally dispatch it
1543 ether_dispatch(ifp, isr, m, cpuid);
1545 logether(pkt_end, ifp);
1549 ether_characterize(struct mbuf **m0)
1551 struct mbuf *m = *m0;
1552 struct ether_header *eh;
1553 uint16_t ether_type;
1556 eh = mtod(m, struct ether_header *);
1557 ether_type = ntohs(eh->ether_type);
1560 * Map ether type to netisr id.
1562 switch (ether_type) {
1574 case ETHERTYPE_IPV6:
1580 case ETHERTYPE_MPLS:
1581 case ETHERTYPE_MPLS_MCAST:
1582 m->m_flags |= M_MPLSLABELED;
1589 * NETISR_MAX is an invalid value; it is chosen to let
1590 * netisr_characterize() know that we have no clear
1591 * idea where this packet should go.
1598 * Ask the isr to characterize the packet since we couldn't.
1599 * This is an attempt to optimally get us onto the correct protocol
1602 netisr_characterize(isr, &m, sizeof(struct ether_header));
1609 ether_demux_handler(netmsg_t nmsg)
1611 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */
1617 ifp = m->m_pkthdr.rcvif;
1619 ether_demux_oncpu(ifp, m);
1623 ether_demux(struct mbuf *m)
1625 struct netmsg_packet *pmsg;
1628 isr = ether_characterize(&m);
1632 KKASSERT(m->m_flags & M_HASH);
1633 pmsg = &m->m_hdr.mh_netmsg;
1634 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1635 0, ether_demux_handler);
1636 pmsg->nm_packet = m;
1637 pmsg->base.lmsg.u.ms_result = isr;
1639 lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg);
1643 kether_aton(const char *macstr, u_char *addr)
1645 unsigned int o0, o1, o2, o3, o4, o5;
1648 if (macstr == NULL || addr == NULL)
1651 n = ksscanf(macstr, "%x:%x:%x:%x:%x:%x", &o0, &o1, &o2,
1667 kether_ntoa(const u_char *addr, char *buf)
1669 int len = ETHER_ADDRSTRLEN + 1;
1672 n = ksnprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x", addr[0],
1673 addr[1], addr[2], addr[3], addr[4], addr[5]);
1681 MODULE_VERSION(ether, 1);