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/dummynet/ip_dummynet.h>
81 #include <netinet6/nd6.h>
85 #include <netinet/ip_carp.h>
89 #include <netproto/mpls/mpls.h>
92 /* netgraph node hooks for ng_ether(4) */
93 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
94 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
95 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
96 void (*ng_ether_attach_p)(struct ifnet *ifp);
97 void (*ng_ether_detach_p)(struct ifnet *ifp);
99 void (*vlan_input_p)(struct mbuf *);
101 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
103 static void ether_restore_header(struct mbuf **, const struct ether_header *,
104 const struct ether_header *);
105 static int ether_characterize(struct mbuf **);
106 static void ether_dispatch(int, struct mbuf *, int);
111 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
112 int (*bridge_output_p)(struct ifnet *, struct mbuf *);
113 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
114 struct ifnet *(*bridge_interface_p)(void *if_bridge);
116 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
119 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
120 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
123 #define gotoerr(e) do { error = (e); goto bad; } while (0)
124 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
126 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
128 const struct ether_header *eh);
130 static int ether_ipfw;
131 static u_long ether_restore_hdr;
132 static u_long ether_prepend_hdr;
133 static u_long ether_input_wronghash;
134 static int ether_debug;
137 static u_long ether_pktinfo_try;
138 static u_long ether_pktinfo_hit;
139 static u_long ether_rss_nopi;
140 static u_long ether_rss_nohash;
141 static u_long ether_input_requeue;
143 static u_long ether_input_wronghwhash;
144 static int ether_input_ckhash;
146 #define ETHER_TSOLEN_DEFAULT (4 * ETHERMTU)
148 static int ether_tsolen_default = ETHER_TSOLEN_DEFAULT;
149 TUNABLE_INT("net.link.ether.tsolen", ðer_tsolen_default);
151 SYSCTL_DECL(_net_link);
152 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
153 SYSCTL_INT(_net_link_ether, OID_AUTO, debug, CTLFLAG_RW,
154 ðer_debug, 0, "Ether debug");
155 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
156 ðer_ipfw, 0, "Pass ether pkts through firewall");
157 SYSCTL_ULONG(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW,
158 ðer_restore_hdr, 0, "# of ether header restoration");
159 SYSCTL_ULONG(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW,
160 ðer_prepend_hdr, 0,
161 "# of ether header restoration which prepends mbuf");
162 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghash, CTLFLAG_RW,
163 ðer_input_wronghash, 0, "# of input packets with wrong hash");
164 SYSCTL_INT(_net_link_ether, OID_AUTO, tsolen, CTLFLAG_RW,
165 ðer_tsolen_default, 0, "Default max TSO length");
168 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nopi, CTLFLAG_RW,
169 ðer_rss_nopi, 0, "# of packets do not have pktinfo");
170 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nohash, CTLFLAG_RW,
171 ðer_rss_nohash, 0, "# of packets do not have hash");
172 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_try, CTLFLAG_RW,
173 ðer_pktinfo_try, 0,
174 "# of tries to find packets' msgport using pktinfo");
175 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_hit, CTLFLAG_RW,
176 ðer_pktinfo_hit, 0,
177 "# of packets whose msgport are found using pktinfo");
178 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_requeue, CTLFLAG_RW,
179 ðer_input_requeue, 0, "# of input packets gets requeued");
181 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghwhash, CTLFLAG_RW,
182 ðer_input_wronghwhash, 0, "# of input packets with wrong hw hash");
183 SYSCTL_INT(_net_link_ether, OID_AUTO, always_ckhash, CTLFLAG_RW,
184 ðer_input_ckhash, 0, "always check hash");
186 #define ETHER_KTR_STR "ifp=%p"
187 #define ETHER_KTR_ARGS struct ifnet *ifp
189 #define KTR_ETHERNET KTR_ALL
191 KTR_INFO_MASTER(ether);
192 KTR_INFO(KTR_ETHERNET, ether, pkt_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARGS);
193 KTR_INFO(KTR_ETHERNET, ether, pkt_end, 1, ETHER_KTR_STR, ETHER_KTR_ARGS);
194 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARGS);
195 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARGS);
196 #define logether(name, arg) KTR_LOG(ether_ ## name, arg)
199 * Ethernet output routine.
200 * Encapsulate a packet of type family for the local net.
201 * Use trailer local net encapsulation if enough data in first
202 * packet leaves a multiple of 512 bytes of data in remainder.
203 * Assumes that ifp is actually pointer to arpcom structure.
206 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
209 struct ether_header *eh, *deh;
212 int hlen = ETHER_HDR_LEN; /* link layer header length */
213 struct arpcom *ac = IFP2AC(ifp);
216 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
218 if (ifp->if_flags & IFF_MONITOR)
220 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
223 M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT);
226 m->m_pkthdr.csum_lhlen = sizeof(struct ether_header);
227 eh = mtod(m, struct ether_header *);
228 edst = eh->ether_dhost;
231 * Fill in the destination ethernet address and frame type.
233 switch (dst->sa_family) {
236 if (!arpresolve(ifp, rt, m, dst, edst))
237 return (0); /* if not yet resolved */
239 if (m->m_flags & M_MPLSLABELED)
240 eh->ether_type = htons(ETHERTYPE_MPLS);
243 eh->ether_type = htons(ETHERTYPE_IP);
248 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst))
249 return (0); /* Something bad happenned. */
250 eh->ether_type = htons(ETHERTYPE_IPV6);
253 case pseudo_AF_HDRCMPLT:
255 loop_copy = -1; /* if this is for us, don't do it */
256 deh = (struct ether_header *)dst->sa_data;
257 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
258 eh->ether_type = deh->ether_type;
262 if_printf(ifp, "can't handle af%d\n", dst->sa_family);
263 gotoerr(EAFNOSUPPORT);
266 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */
267 memcpy(eh->ether_shost,
268 ((struct ether_header *)dst->sa_data)->ether_shost,
271 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
274 * Bridges require special output handling.
276 if (ifp->if_bridge) {
277 KASSERT(bridge_output_p != NULL,
278 ("%s: if_bridge not loaded!", __func__));
279 return bridge_output_p(ifp, m);
283 * If a simplex interface, and the packet is being sent to our
284 * Ethernet address or a broadcast address, loopback a copy.
285 * XXX To make a simplex device behave exactly like a duplex
286 * device, we should copy in the case of sending to our own
287 * ethernet address (thus letting the original actually appear
288 * on the wire). However, we don't do that here for security
289 * reasons and compatibility with the original behavior.
291 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
294 if (m->m_pkthdr.csum_flags & CSUM_IP)
295 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
296 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
297 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
298 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
301 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) {
302 n->m_pkthdr.csum_flags |= csum_flags;
303 if (csum_flags & CSUM_DATA_VALID)
304 n->m_pkthdr.csum_data = 0xffff;
305 if_simloop(ifp, n, dst->sa_family, hlen);
307 IFNET_STAT_INC(ifp, iqdrops, 1);
308 } else if (bcmp(eh->ether_dhost, eh->ether_shost,
309 ETHER_ADDR_LEN) == 0) {
310 m->m_pkthdr.csum_flags |= csum_flags;
311 if (csum_flags & CSUM_DATA_VALID)
312 m->m_pkthdr.csum_data = 0xffff;
313 if_simloop(ifp, m, dst->sa_family, hlen);
314 return (0); /* XXX */
319 if (ifp->if_type == IFT_CARP) {
320 ifp = carp_parent(ifp);
322 gotoerr(ENETUNREACH);
327 * Check precondition again
329 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
331 if (ifp->if_flags & IFF_MONITOR)
333 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
334 (IFF_UP | IFF_RUNNING))
339 /* Handle ng_ether(4) processing, if any */
340 if (ng_ether_output_p != NULL) {
342 * Hold BGL and recheck ng_ether_output_p
345 if (ng_ether_output_p != NULL) {
346 if ((error = ng_ether_output_p(ifp, &m)) != 0) {
358 /* Continue with link-layer output */
359 return ether_output_frame(ifp, m);
367 * Returns the bridge interface an ifp is associated
370 * Only call if ifp->if_bridge != NULL.
373 ether_bridge_interface(struct ifnet *ifp)
375 if (bridge_interface_p)
376 return(bridge_interface_p(ifp->if_bridge));
381 * Ethernet link layer output routine to send a raw frame to the device.
383 * This assumes that the 14 byte Ethernet header is present and contiguous
387 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
389 struct ip_fw *rule = NULL;
391 struct altq_pktattr pktattr;
393 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
395 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
398 /* Extract info from dummynet tag */
399 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
400 KKASSERT(mtag != NULL);
401 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
402 KKASSERT(rule != NULL);
404 m_tag_delete(m, mtag);
405 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
408 if (ifq_is_enabled(&ifp->if_snd))
409 altq_etherclassify(&ifp->if_snd, m, &pktattr);
411 if (IPFW_LOADED && ether_ipfw != 0) {
412 struct ether_header save_eh, *eh;
414 eh = mtod(m, struct ether_header *);
416 m_adj(m, ETHER_HDR_LEN);
417 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) {
421 return ENOBUFS; /* pkt dropped */
423 return 0; /* consumed e.g. in a pipe */
426 /* packet was ok, restore the ethernet header */
427 ether_restore_header(&m, eh, &save_eh);
436 * Queue message on interface, update output statistics if
437 * successful, and start output if interface not yet active.
439 error = ifq_dispatch(ifp, m, &pktattr);
444 * ipfw processing for ethernet packets (in and out).
445 * The second parameter is NULL from ether_demux(), and ifp from
446 * ether_output_frame().
449 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule,
450 const struct ether_header *eh)
452 struct ether_header save_eh = *eh; /* might be a ptr in *m0 */
453 struct ip_fw_args args;
458 if (*rule != NULL && fw_one_pass)
459 return TRUE; /* dummynet packet, already partially processed */
462 * I need some amount of data to be contiguous.
464 i = min((*m0)->m_pkthdr.len, max_protohdr);
465 if ((*m0)->m_len < i) {
466 *m0 = m_pullup(*m0, i);
474 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
475 m_tag_delete(*m0, mtag);
476 if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
477 mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL);
478 KKASSERT(mtag != NULL);
479 m_tag_delete(*m0, mtag);
480 (*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED;
483 args.m = *m0; /* the packet we are looking at */
484 args.oif = dst; /* destination, if any */
485 args.rule = *rule; /* matching rule to restart */
486 args.eh = &save_eh; /* MAC header for bridged/MAC packets */
487 i = ip_fw_chk_ptr(&args);
502 * XXX at some point add support for divert/forward actions.
503 * If none of the above matches, we have to drop the pkt.
509 * Pass the pkt to dummynet, which consumes it.
511 m = *m0; /* pass the original to dummynet */
512 *m0 = NULL; /* and nothing back to the caller */
514 ether_restore_header(&m, eh, &save_eh);
518 ip_fw_dn_io_ptr(m, args.cookie,
519 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
524 panic("unknown ipfw return value: %d", i);
529 ether_input(struct ifnet *ifp, struct mbuf *m)
531 ether_input_pkt(ifp, m, NULL, -1);
535 * Perform common duties while attaching to interface list
538 ether_ifattach(struct ifnet *ifp, const uint8_t *lla,
539 lwkt_serialize_t serializer)
541 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
546 ether_ifattach_bpf(struct ifnet *ifp, const uint8_t *lla,
547 u_int dlt, u_int hdrlen, lwkt_serialize_t serializer)
549 struct sockaddr_dl *sdl;
550 char ethstr[ETHER_ADDRSTRLEN + 1];
554 ifp->if_type = IFT_ETHER;
555 ifp->if_addrlen = ETHER_ADDR_LEN;
556 ifp->if_hdrlen = ETHER_HDR_LEN;
557 if_attach(ifp, serializer);
559 for (i = 0; i < ifq->altq_subq_cnt; ++i) {
560 struct ifaltq_subque *ifsq = ifq_get_subq(ifq, i);
562 ifsq->ifsq_maxbcnt = ifsq->ifsq_maxlen *
563 (ETHER_MAX_LEN - ETHER_CRC_LEN);
565 ifp->if_mtu = ETHERMTU;
566 if (ifp->if_tsolen <= 0) {
567 if ((ether_tsolen_default / ETHERMTU) < 2) {
568 kprintf("ether TSO maxlen %d -> %d\n",
569 ether_tsolen_default, ETHER_TSOLEN_DEFAULT);
570 ether_tsolen_default = ETHER_TSOLEN_DEFAULT;
572 ifp->if_tsolen = ether_tsolen_default;
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: %s\n", kether_ntoa(lla, ethstr));
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, u_long 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_IFNET_SERIALIZED_ALL(ifp);
630 switch (ifa->ifa_addr->sa_family) {
633 IF_INIT(ifp); /* before arpwhohas */
634 arp_ifinit(ifp, ifa);
644 bcopy(IFP2AC(ifp)->ac_enaddr,
645 ((struct sockaddr *)ifr->ifr_data)->sa_data,
651 * Set the interface MTU.
653 if (ifr->ifr_mtu > ETHERMTU) {
656 ifp->if_mtu = ifr->ifr_mtu;
671 struct sockaddr **llsa,
674 struct sockaddr_dl *sdl;
676 struct sockaddr_in *sin;
679 struct sockaddr_in6 *sin6;
683 switch(sa->sa_family) {
686 * No mapping needed. Just check that it's a valid MC address.
688 sdl = (struct sockaddr_dl *)sa;
689 e_addr = LLADDR(sdl);
690 if ((e_addr[0] & 1) != 1)
691 return EADDRNOTAVAIL;
697 sin = (struct sockaddr_in *)sa;
698 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
699 return EADDRNOTAVAIL;
700 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
701 sdl->sdl_len = sizeof *sdl;
702 sdl->sdl_family = AF_LINK;
703 sdl->sdl_index = ifp->if_index;
704 sdl->sdl_type = IFT_ETHER;
705 sdl->sdl_alen = ETHER_ADDR_LEN;
706 e_addr = LLADDR(sdl);
707 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
708 *llsa = (struct sockaddr *)sdl;
713 sin6 = (struct sockaddr_in6 *)sa;
714 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
716 * An IP6 address of 0 means listen to all
717 * of the Ethernet multicast address used for IP6.
718 * (This is used for multicast routers.)
720 ifp->if_flags |= IFF_ALLMULTI;
724 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
725 return EADDRNOTAVAIL;
726 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
727 sdl->sdl_len = sizeof *sdl;
728 sdl->sdl_family = AF_LINK;
729 sdl->sdl_index = ifp->if_index;
730 sdl->sdl_type = IFT_ETHER;
731 sdl->sdl_alen = ETHER_ADDR_LEN;
732 e_addr = LLADDR(sdl);
733 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
734 *llsa = (struct sockaddr *)sdl;
740 * Well, the text isn't quite right, but it's the name
749 * This is for reference. We have a table-driven version
750 * of the little-endian crc32 generator, which is faster
751 * than the double-loop.
754 ether_crc32_le(const uint8_t *buf, size_t len)
756 uint32_t c, crc, carry;
759 crc = 0xffffffffU; /* initial value */
761 for (i = 0; i < len; i++) {
763 for (j = 0; j < 8; j++) {
764 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
768 crc = (crc ^ ETHER_CRC_POLY_LE);
776 ether_crc32_le(const uint8_t *buf, size_t len)
778 static const uint32_t crctab[] = {
779 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
780 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
781 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
782 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
787 crc = 0xffffffffU; /* initial value */
789 for (i = 0; i < len; i++) {
791 crc = (crc >> 4) ^ crctab[crc & 0xf];
792 crc = (crc >> 4) ^ crctab[crc & 0xf];
800 ether_crc32_be(const uint8_t *buf, size_t len)
802 uint32_t c, crc, carry;
805 crc = 0xffffffffU; /* initial value */
807 for (i = 0; i < len; i++) {
809 for (j = 0; j < 8; j++) {
810 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
814 crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
822 * find the size of ethernet header, and call classifier
825 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
826 struct altq_pktattr *pktattr)
828 struct ether_header *eh;
830 int hlen, af, hdrsize;
832 hlen = sizeof(struct ether_header);
833 eh = mtod(m, struct ether_header *);
835 ether_type = ntohs(eh->ether_type);
836 if (ether_type < ETHERMTU) {
838 struct llc *llc = (struct llc *)(eh + 1);
841 if (m->m_len < hlen ||
842 llc->llc_dsap != LLC_SNAP_LSAP ||
843 llc->llc_ssap != LLC_SNAP_LSAP ||
844 llc->llc_control != LLC_UI)
845 goto bad; /* not snap! */
847 ether_type = ntohs(llc->llc_un.type_snap.ether_type);
850 if (ether_type == ETHERTYPE_IP) {
852 hdrsize = 20; /* sizeof(struct ip) */
854 } else if (ether_type == ETHERTYPE_IPV6) {
856 hdrsize = 40; /* sizeof(struct ip6_hdr) */
861 while (m->m_len <= hlen) {
865 if (m->m_len < hlen + hdrsize) {
867 * ip header is not in a single mbuf. this should not
868 * happen in the current code.
869 * (todo: use m_pulldown in the future)
875 ifq_classify(ifq, m, af, pktattr);
882 pktattr->pattr_class = NULL;
883 pktattr->pattr_hdr = NULL;
884 pktattr->pattr_af = AF_UNSPEC;
888 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
889 const struct ether_header *save_eh)
891 struct mbuf *m = *m0;
896 * Prepend the header, optimize for the common case of
897 * eh pointing into the mbuf.
899 if ((const void *)(eh + 1) == (void *)m->m_data) {
900 m->m_data -= ETHER_HDR_LEN;
901 m->m_len += ETHER_HDR_LEN;
902 m->m_pkthdr.len += ETHER_HDR_LEN;
906 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
908 bcopy(save_eh, mtod(m, struct ether_header *),
916 * Upper layer processing for a received Ethernet packet.
919 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
921 struct ether_header *eh;
922 int isr, discard = 0;
924 struct ip_fw *rule = NULL;
927 KASSERT(m->m_len >= ETHER_HDR_LEN,
928 ("ether header is not contiguous!"));
930 eh = mtod(m, struct ether_header *);
932 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
935 /* Extract info from dummynet tag */
936 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
937 KKASSERT(mtag != NULL);
938 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
939 KKASSERT(rule != NULL);
941 m_tag_delete(m, mtag);
942 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
944 /* packet is passing the second time */
949 * We got a packet which was unicast to a different Ethernet
950 * address. If the driver is working properly, then this
951 * situation can only happen when the interface is in
952 * promiscuous mode. We defer the packet discarding until the
953 * vlan processing is done, so that vlan/bridge or vlan/netgraph
956 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
957 !ETHER_IS_MULTICAST(eh->ether_dhost) &&
958 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
959 if (ether_debug & 1) {
960 kprintf("%02x:%02x:%02x:%02x:%02x:%02x "
961 "%02x:%02x:%02x:%02x:%02x:%02x "
962 "%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n",
976 ((u_char *)IFP2AC(ifp)->ac_enaddr)[0],
977 ((u_char *)IFP2AC(ifp)->ac_enaddr)[1],
978 ((u_char *)IFP2AC(ifp)->ac_enaddr)[2],
979 ((u_char *)IFP2AC(ifp)->ac_enaddr)[3],
980 ((u_char *)IFP2AC(ifp)->ac_enaddr)[4],
981 ((u_char *)IFP2AC(ifp)->ac_enaddr)[5]
984 if ((ether_debug & 2) == 0)
989 if (IPFW_LOADED && ether_ipfw != 0 && !discard) {
990 struct ether_header save_eh = *eh;
992 /* XXX old crufty stuff, needs to be removed */
993 m_adj(m, sizeof(struct ether_header));
995 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
1000 ether_restore_header(&m, eh, &save_eh);
1003 eh = mtod(m, struct ether_header *);
1006 ether_type = ntohs(eh->ether_type);
1007 KKASSERT(ether_type != ETHERTYPE_VLAN);
1009 if (m->m_flags & M_VLANTAG) {
1010 void (*vlan_input_func)(struct mbuf *);
1012 vlan_input_func = vlan_input_p;
1013 if (vlan_input_func != NULL) {
1016 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1023 * If we have been asked to discard this packet
1024 * (e.g. not for us), drop it before entering
1033 * Clear protocol specific flags,
1034 * before entering the upper layer.
1036 m->m_flags &= ~M_ETHER_FLAGS;
1038 /* Strip ethernet header. */
1039 m_adj(m, sizeof(struct ether_header));
1041 switch (ether_type) {
1044 if ((m->m_flags & M_LENCHECKED) == 0) {
1045 if (!ip_lengthcheck(&m, 0))
1048 if (ipflow_fastforward(m))
1054 if (ifp->if_flags & IFF_NOARP) {
1055 /* Discard packet if ARP is disabled on interface */
1064 case ETHERTYPE_IPV6:
1070 case ETHERTYPE_MPLS:
1071 case ETHERTYPE_MPLS_MCAST:
1072 /* Should have been set by ether_input_pkt(). */
1073 KKASSERT(m->m_flags & M_MPLSLABELED);
1080 * The accurate msgport is not determined before
1081 * we reach here, so recharacterize packet.
1083 m->m_flags &= ~M_HASH;
1084 if (ng_ether_input_orphan_p != NULL) {
1086 * Put back the ethernet header so netgraph has a
1087 * consistent view of inbound packets.
1089 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
1092 * M_PREPEND frees the mbuf in case of failure.
1097 * Hold BGL and recheck ng_ether_input_orphan_p
1100 if (ng_ether_input_orphan_p != NULL) {
1101 ng_ether_input_orphan_p(ifp, m);
1111 if (m->m_flags & M_HASH) {
1112 if (&curthread->td_msgport ==
1113 netisr_hashport(m->m_pkthdr.hash)) {
1114 netisr_handle(isr, m);
1118 * XXX Something is wrong,
1119 * we probably should panic here!
1121 m->m_flags &= ~M_HASH;
1122 atomic_add_long(ðer_input_wronghash, 1);
1126 atomic_add_long(ðer_input_requeue, 1);
1128 netisr_queue(isr, m);
1132 * First we perform any link layer operations, then continue to the
1133 * upper layers with ether_demux_oncpu().
1136 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1142 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1144 * Receiving interface's flags are changed, when this
1145 * packet is waiting for processing; discard it.
1152 * Tap the packet off here for a bridge. bridge_input()
1153 * will return NULL if it has consumed the packet, otherwise
1154 * it gets processed as normal. Note that bridge_input()
1155 * will always return the original packet if we need to
1156 * process it locally.
1158 if (ifp->if_bridge) {
1159 KASSERT(bridge_input_p != NULL,
1160 ("%s: if_bridge not loaded!", __func__));
1162 if(m->m_flags & M_ETHER_BRIDGED) {
1163 m->m_flags &= ~M_ETHER_BRIDGED;
1165 m = bridge_input_p(ifp, m);
1169 KASSERT(ifp == m->m_pkthdr.rcvif,
1170 ("bridge_input_p changed rcvif"));
1175 carp = ifp->if_carp;
1177 m = carp_input(carp, m);
1180 KASSERT(ifp == m->m_pkthdr.rcvif,
1181 ("carp_input changed rcvif"));
1185 /* Handle ng_ether(4) processing, if any */
1186 if (ng_ether_input_p != NULL) {
1188 * Hold BGL and recheck ng_ether_input_p
1191 if (ng_ether_input_p != NULL)
1192 ng_ether_input_p(ifp, &m);
1199 /* Continue with upper layer processing */
1200 ether_demux_oncpu(ifp, m);
1204 * Perform certain functions of ether_input_pkt():
1206 * - Update statistics
1207 * - Run bpf(4) tap if requested
1208 * Then pass the packet to ether_input_oncpu().
1210 * This function should be used by pseudo interface (e.g. vlan(4)),
1211 * when it tries to claim that the packet is received by it.
1217 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags)
1219 /* Discard packet if interface is not up */
1220 if (!(ifp->if_flags & IFF_UP)) {
1226 * Change receiving interface. The bridge will often pass a flag to
1227 * ask that this not be done so ARPs get applied to the correct
1230 if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 ||
1231 m->m_pkthdr.rcvif == NULL) {
1232 m->m_pkthdr.rcvif = ifp;
1235 /* Update statistics */
1236 IFNET_STAT_INC(ifp, ipackets, 1);
1237 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1238 if (m->m_flags & (M_MCAST | M_BCAST))
1239 IFNET_STAT_INC(ifp, imcasts, 1);
1241 if (reinput_flags & REINPUT_RUNBPF)
1244 ether_input_oncpu(ifp, m);
1247 static __inline boolean_t
1248 ether_vlancheck(struct mbuf **m0)
1250 struct mbuf *m = *m0;
1251 struct ether_header *eh;
1252 uint16_t ether_type;
1254 eh = mtod(m, struct ether_header *);
1255 ether_type = ntohs(eh->ether_type);
1257 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) {
1259 * Extract vlan tag if hardware does not do it for us
1261 vlan_ether_decap(&m);
1265 eh = mtod(m, struct ether_header *);
1266 ether_type = ntohs(eh->ether_type);
1269 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) {
1271 * To prevent possible dangerous recursion,
1272 * we don't do vlan-in-vlan
1274 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1277 KKASSERT(ether_type != ETHERTYPE_VLAN);
1279 m->m_flags |= M_ETHER_VLANCHECKED;
1290 ether_input_handler(netmsg_t nmsg)
1292 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */
1293 struct ether_header *eh;
1300 if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) {
1301 if (!ether_vlancheck(&m)) {
1302 KKASSERT(m == NULL);
1306 if ((m->m_flags & (M_HASH | M_CKHASH)) == (M_HASH | M_CKHASH) ||
1307 __predict_false(ether_input_ckhash)) {
1311 * Need to verify the hash supplied by the hardware
1312 * which could be wrong.
1314 m->m_flags &= ~(M_HASH | M_CKHASH);
1315 isr = ether_characterize(&m);
1318 KKASSERT(m->m_flags & M_HASH);
1320 if (netisr_hashcpu(m->m_pkthdr.hash) != mycpuid) {
1322 * Wrong hardware supplied hash; redispatch
1324 ether_dispatch(isr, m, -1);
1325 if (__predict_false(ether_input_ckhash))
1326 atomic_add_long(ðer_input_wronghwhash, 1);
1330 ifp = m->m_pkthdr.rcvif;
1332 eh = mtod(m, struct ether_header *);
1333 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1334 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1335 ifp->if_addrlen) == 0)
1336 m->m_flags |= M_BCAST;
1338 m->m_flags |= M_MCAST;
1339 IFNET_STAT_INC(ifp, imcasts, 1);
1342 ether_input_oncpu(ifp, m);
1346 * Send the packet to the target netisr msgport
1348 * At this point the packet must be characterized (M_HASH set),
1349 * so we know which netisr to send it to.
1352 ether_dispatch(int isr, struct mbuf *m, int cpuid)
1354 struct netmsg_packet *pmsg;
1357 KKASSERT(m->m_flags & M_HASH);
1358 target_cpuid = netisr_hashcpu(m->m_pkthdr.hash);
1360 pmsg = &m->m_hdr.mh_netmsg;
1361 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1362 0, ether_input_handler);
1363 pmsg->nm_packet = m;
1364 pmsg->base.lmsg.u.ms_result = isr;
1366 logether(disp_beg, NULL);
1367 if (target_cpuid == cpuid) {
1368 lwkt_sendmsg_oncpu(netisr_cpuport(target_cpuid),
1371 lwkt_sendmsg(netisr_cpuport(target_cpuid),
1374 logether(disp_end, NULL);
1378 * Process a received Ethernet packet.
1380 * The ethernet header is assumed to be in the mbuf so the caller
1381 * MUST MAKE SURE that there are at least sizeof(struct ether_header)
1382 * bytes in the first mbuf.
1384 * If the caller knows that the current thread is stick to the current
1385 * cpu, e.g. the interrupt thread or the netisr thread, the current cpuid
1386 * (mycpuid) should be passed through 'cpuid' argument. Else -1 should
1387 * be passed as 'cpuid' argument.
1390 ether_input_pkt(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi,
1397 /* Discard packet if interface is not up */
1398 if (!(ifp->if_flags & IFF_UP)) {
1403 if (m->m_len < sizeof(struct ether_header)) {
1404 /* XXX error in the caller. */
1409 m->m_pkthdr.rcvif = ifp;
1411 logether(pkt_beg, ifp);
1413 ETHER_BPF_MTAP(ifp, m);
1415 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1417 if (ifp->if_flags & IFF_MONITOR) {
1418 struct ether_header *eh;
1420 eh = mtod(m, struct ether_header *);
1421 if (ETHER_IS_MULTICAST(eh->ether_dhost))
1422 IFNET_STAT_INC(ifp, imcasts, 1);
1425 * Interface marked for monitoring; discard packet.
1429 logether(pkt_end, ifp);
1434 * If the packet has been characterized (pi->pi_netisr / M_HASH)
1435 * we can dispatch it immediately with trivial checks.
1437 if (pi != NULL && (m->m_flags & M_HASH)) {
1439 atomic_add_long(ðer_pktinfo_try, 1);
1441 netisr_hashcheck(pi->pi_netisr, m, pi);
1442 if (m->m_flags & M_HASH) {
1443 ether_dispatch(pi->pi_netisr, m, cpuid);
1445 atomic_add_long(ðer_pktinfo_hit, 1);
1447 logether(pkt_end, ifp);
1452 else if (ifp->if_capenable & IFCAP_RSS) {
1454 atomic_add_long(ðer_rss_nopi, 1);
1456 atomic_add_long(ðer_rss_nohash, 1);
1461 * Packet hash will be recalculated by software, so clear
1462 * the M_HASH and M_CKHASH flag set by the driver; the hash
1463 * value calculated by the hardware may not be exactly what
1466 m->m_flags &= ~(M_HASH | M_CKHASH);
1468 if (!ether_vlancheck(&m)) {
1469 KKASSERT(m == NULL);
1470 logether(pkt_end, ifp);
1474 isr = ether_characterize(&m);
1476 logether(pkt_end, ifp);
1481 * Finally dispatch it
1483 ether_dispatch(isr, m, cpuid);
1485 logether(pkt_end, ifp);
1489 ether_characterize(struct mbuf **m0)
1491 struct mbuf *m = *m0;
1492 struct ether_header *eh;
1493 uint16_t ether_type;
1496 eh = mtod(m, struct ether_header *);
1497 ether_type = ntohs(eh->ether_type);
1500 * Map ether type to netisr id.
1502 switch (ether_type) {
1514 case ETHERTYPE_IPV6:
1520 case ETHERTYPE_MPLS:
1521 case ETHERTYPE_MPLS_MCAST:
1522 m->m_flags |= M_MPLSLABELED;
1529 * NETISR_MAX is an invalid value; it is chosen to let
1530 * netisr_characterize() know that we have no clear
1531 * idea where this packet should go.
1538 * Ask the isr to characterize the packet since we couldn't.
1539 * This is an attempt to optimally get us onto the correct protocol
1542 netisr_characterize(isr, &m, sizeof(struct ether_header));
1549 ether_demux_handler(netmsg_t nmsg)
1551 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */
1557 ifp = m->m_pkthdr.rcvif;
1559 ether_demux_oncpu(ifp, m);
1563 ether_demux(struct mbuf *m)
1565 struct netmsg_packet *pmsg;
1568 isr = ether_characterize(&m);
1572 KKASSERT(m->m_flags & M_HASH);
1573 pmsg = &m->m_hdr.mh_netmsg;
1574 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1575 0, ether_demux_handler);
1576 pmsg->nm_packet = m;
1577 pmsg->base.lmsg.u.ms_result = isr;
1579 lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg);
1583 kether_aton(const char *macstr, u_char *addr)
1585 unsigned int o0, o1, o2, o3, o4, o5;
1588 if (macstr == NULL || addr == NULL)
1591 n = ksscanf(macstr, "%x:%x:%x:%x:%x:%x", &o0, &o1, &o2,
1607 kether_ntoa(const u_char *addr, char *buf)
1609 int len = ETHER_ADDRSTRLEN + 1;
1612 n = ksnprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x", addr[0],
1613 addr[1], addr[2], addr[3], addr[4], addr[5]);
1621 MODULE_VERSION(ether, 1);