3 * Linux ethernet bridge
6 * Lennert Buytenhek <buytenh@gnu.org>
7 * Bart De Schuymer <bdschuym@pandora.be>
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
14 * Lennert dedicates this file to Kerstin Wurdinger.
17 #include <linux/module.h>
18 #include <linux/kernel.h>
19 #include <linux/slab.h>
21 #include <linux/netdevice.h>
22 #include <linux/skbuff.h>
23 #include <linux/if_arp.h>
24 #include <linux/if_ether.h>
25 #include <linux/if_vlan.h>
26 #include <linux/if_pppox.h>
27 #include <linux/ppp_defs.h>
28 #include <linux/netfilter_bridge.h>
29 #include <linux/netfilter_ipv4.h>
30 #include <linux/netfilter_ipv6.h>
31 #include <linux/netfilter_arp.h>
32 #include <linux/in_route.h>
33 #include <linux/rculist.h>
34 #include <linux/inetdevice.h>
38 #include <net/addrconf.h>
39 #include <net/route.h>
40 #include <net/netfilter/br_netfilter.h>
41 #include <net/netns/generic.h>
43 #include <asm/uaccess.h>
44 #include "br_private.h"
46 #include <linux/sysctl.h>
49 static int brnf_net_id __read_mostly;
56 static struct ctl_table_header *brnf_sysctl_header;
57 static int brnf_call_iptables __read_mostly = 1;
58 static int brnf_call_ip6tables __read_mostly = 1;
59 static int brnf_call_arptables __read_mostly = 1;
60 static int brnf_filter_vlan_tagged __read_mostly;
61 static int brnf_filter_pppoe_tagged __read_mostly;
62 static int brnf_pass_vlan_indev __read_mostly;
64 #define brnf_call_iptables 1
65 #define brnf_call_ip6tables 1
66 #define brnf_call_arptables 1
67 #define brnf_filter_vlan_tagged 0
68 #define brnf_filter_pppoe_tagged 0
69 #define brnf_pass_vlan_indev 0
73 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP))
75 #define IS_IPV6(skb) \
76 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6))
79 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP))
81 static inline __be16 vlan_proto(const struct sk_buff *skb)
83 if (skb_vlan_tag_present(skb))
85 else if (skb->protocol == htons(ETH_P_8021Q))
86 return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
91 #define IS_VLAN_IP(skb) \
92 (vlan_proto(skb) == htons(ETH_P_IP) && \
93 brnf_filter_vlan_tagged)
95 #define IS_VLAN_IPV6(skb) \
96 (vlan_proto(skb) == htons(ETH_P_IPV6) && \
97 brnf_filter_vlan_tagged)
99 #define IS_VLAN_ARP(skb) \
100 (vlan_proto(skb) == htons(ETH_P_ARP) && \
101 brnf_filter_vlan_tagged)
103 static inline __be16 pppoe_proto(const struct sk_buff *skb)
105 return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
106 sizeof(struct pppoe_hdr)));
109 #define IS_PPPOE_IP(skb) \
110 (skb->protocol == htons(ETH_P_PPP_SES) && \
111 pppoe_proto(skb) == htons(PPP_IP) && \
112 brnf_filter_pppoe_tagged)
114 #define IS_PPPOE_IPV6(skb) \
115 (skb->protocol == htons(ETH_P_PPP_SES) && \
116 pppoe_proto(skb) == htons(PPP_IPV6) && \
117 brnf_filter_pppoe_tagged)
119 /* largest possible L2 header, see br_nf_dev_queue_xmit() */
120 #define NF_BRIDGE_MAX_MAC_HEADER_LENGTH (PPPOE_SES_HLEN + ETH_HLEN)
122 struct brnf_frag_data {
123 char mac[NF_BRIDGE_MAX_MAC_HEADER_LENGTH];
130 static DEFINE_PER_CPU(struct brnf_frag_data, brnf_frag_data_storage);
132 static void nf_bridge_info_free(struct sk_buff *skb)
134 if (skb->nf_bridge) {
135 nf_bridge_put(skb->nf_bridge);
136 skb->nf_bridge = NULL;
140 static inline struct net_device *bridge_parent(const struct net_device *dev)
142 struct net_bridge_port *port;
144 port = br_port_get_rcu(dev);
145 return port ? port->br->dev : NULL;
148 static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
150 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
152 if (atomic_read(&nf_bridge->use) > 1) {
153 struct nf_bridge_info *tmp = nf_bridge_alloc(skb);
156 memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info));
157 atomic_set(&tmp->use, 1);
159 nf_bridge_put(nf_bridge);
165 unsigned int nf_bridge_encap_header_len(const struct sk_buff *skb)
167 switch (skb->protocol) {
168 case __cpu_to_be16(ETH_P_8021Q):
170 case __cpu_to_be16(ETH_P_PPP_SES):
171 return PPPOE_SES_HLEN;
177 static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
179 unsigned int len = nf_bridge_encap_header_len(skb);
182 skb->network_header += len;
185 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
187 unsigned int len = nf_bridge_encap_header_len(skb);
189 skb_pull_rcsum(skb, len);
190 skb->network_header += len;
193 /* When handing a packet over to the IP layer
194 * check whether we have a skb that is in the
198 static int br_validate_ipv4(struct net *net, struct sk_buff *skb)
200 const struct iphdr *iph;
203 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
208 /* Basic sanity checks */
209 if (iph->ihl < 5 || iph->version != 4)
212 if (!pskb_may_pull(skb, iph->ihl*4))
216 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
219 len = ntohs(iph->tot_len);
220 if (skb->len < len) {
221 __IP_INC_STATS(net, IPSTATS_MIB_INTRUNCATEDPKTS);
223 } else if (len < (iph->ihl*4))
226 if (pskb_trim_rcsum(skb, len)) {
227 __IP_INC_STATS(net, IPSTATS_MIB_INDISCARDS);
231 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
232 /* We should really parse IP options here but until
233 * somebody who actually uses IP options complains to
234 * us we'll just silently ignore the options because
240 __IP_INC_STATS(net, IPSTATS_MIB_INHDRERRORS);
245 void nf_bridge_update_protocol(struct sk_buff *skb)
247 switch (skb->nf_bridge->orig_proto) {
248 case BRNF_PROTO_8021Q:
249 skb->protocol = htons(ETH_P_8021Q);
251 case BRNF_PROTO_PPPOE:
252 skb->protocol = htons(ETH_P_PPP_SES);
254 case BRNF_PROTO_UNCHANGED:
259 /* Obtain the correct destination MAC address, while preserving the original
260 * source MAC address. If we already know this address, we just copy it. If we
261 * don't, we use the neighbour framework to find out. In both cases, we make
262 * sure that br_handle_frame_finish() is called afterwards.
264 int br_nf_pre_routing_finish_bridge(struct net *net, struct sock *sk, struct sk_buff *skb)
266 struct neighbour *neigh;
267 struct dst_entry *dst;
269 skb->dev = bridge_parent(skb->dev);
273 neigh = dst_neigh_lookup_skb(dst, skb);
275 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
278 if ((neigh->nud_state & NUD_CONNECTED) && neigh->hh.hh_len) {
279 neigh_hh_bridge(&neigh->hh, skb);
280 skb->dev = nf_bridge->physindev;
281 ret = br_handle_frame_finish(net, sk, skb);
283 /* the neighbour function below overwrites the complete
284 * MAC header, so we save the Ethernet source address and
287 skb_copy_from_linear_data_offset(skb,
288 -(ETH_HLEN-ETH_ALEN),
289 nf_bridge->neigh_header,
291 /* tell br_dev_xmit to continue with forwarding */
292 nf_bridge->bridged_dnat = 1;
293 /* FIXME Need to refragment */
294 ret = neigh->output(neigh, skb);
296 neigh_release(neigh);
305 br_nf_ipv4_daddr_was_changed(const struct sk_buff *skb,
306 const struct nf_bridge_info *nf_bridge)
308 return ip_hdr(skb)->daddr != nf_bridge->ipv4_daddr;
311 /* This requires some explaining. If DNAT has taken place,
312 * we will need to fix up the destination Ethernet address.
313 * This is also true when SNAT takes place (for the reply direction).
315 * There are two cases to consider:
316 * 1. The packet was DNAT'ed to a device in the same bridge
317 * port group as it was received on. We can still bridge
319 * 2. The packet was DNAT'ed to a different device, either
320 * a non-bridged device or another bridge port group.
321 * The packet will need to be routed.
323 * The correct way of distinguishing between these two cases is to
324 * call ip_route_input() and to look at skb->dst->dev, which is
325 * changed to the destination device if ip_route_input() succeeds.
327 * Let's first consider the case that ip_route_input() succeeds:
329 * If the output device equals the logical bridge device the packet
330 * came in on, we can consider this bridging. The corresponding MAC
331 * address will be obtained in br_nf_pre_routing_finish_bridge.
332 * Otherwise, the packet is considered to be routed and we just
333 * change the destination MAC address so that the packet will
334 * later be passed up to the IP stack to be routed. For a redirected
335 * packet, ip_route_input() will give back the localhost as output device,
336 * which differs from the bridge device.
338 * Let's now consider the case that ip_route_input() fails:
340 * This can be because the destination address is martian, in which case
341 * the packet will be dropped.
342 * If IP forwarding is disabled, ip_route_input() will fail, while
343 * ip_route_output_key() can return success. The source
344 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
345 * thinks we're handling a locally generated packet and won't care
346 * if IP forwarding is enabled. If the output device equals the logical bridge
347 * device, we proceed as if ip_route_input() succeeded. If it differs from the
348 * logical bridge port or if ip_route_output_key() fails we drop the packet.
350 static int br_nf_pre_routing_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
352 struct net_device *dev = skb->dev;
353 struct iphdr *iph = ip_hdr(skb);
354 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
358 nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
360 if (nf_bridge->pkt_otherhost) {
361 skb->pkt_type = PACKET_OTHERHOST;
362 nf_bridge->pkt_otherhost = false;
364 nf_bridge->in_prerouting = 0;
365 if (br_nf_ipv4_daddr_was_changed(skb, nf_bridge)) {
366 if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
367 struct in_device *in_dev = __in_dev_get_rcu(dev);
369 /* If err equals -EHOSTUNREACH the error is due to a
370 * martian destination or due to the fact that
371 * forwarding is disabled. For most martian packets,
372 * ip_route_output_key() will fail. It won't fail for 2 types of
373 * martian destinations: loopback destinations and destination
374 * 0.0.0.0. In both cases the packet will be dropped because the
375 * destination is the loopback device and not the bridge. */
376 if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
379 rt = ip_route_output(net, iph->daddr, 0,
380 RT_TOS(iph->tos), 0);
382 /* - Bridged-and-DNAT'ed traffic doesn't
383 * require ip_forwarding. */
384 if (rt->dst.dev == dev) {
385 skb_dst_set(skb, &rt->dst);
394 if (skb_dst(skb)->dev == dev) {
396 skb->dev = nf_bridge->physindev;
397 nf_bridge_update_protocol(skb);
398 nf_bridge_push_encap_header(skb);
399 br_nf_hook_thresh(NF_BR_PRE_ROUTING,
400 net, sk, skb, skb->dev,
402 br_nf_pre_routing_finish_bridge);
405 ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr);
406 skb->pkt_type = PACKET_HOST;
409 rt = bridge_parent_rtable(nf_bridge->physindev);
414 skb_dst_set_noref(skb, &rt->dst);
417 skb->dev = nf_bridge->physindev;
418 nf_bridge_update_protocol(skb);
419 nf_bridge_push_encap_header(skb);
420 br_nf_hook_thresh(NF_BR_PRE_ROUTING, net, sk, skb, skb->dev, NULL,
421 br_handle_frame_finish);
425 static struct net_device *brnf_get_logical_dev(struct sk_buff *skb, const struct net_device *dev)
427 struct net_device *vlan, *br;
429 br = bridge_parent(dev);
430 if (brnf_pass_vlan_indev == 0 || !skb_vlan_tag_present(skb))
433 vlan = __vlan_find_dev_deep_rcu(br, skb->vlan_proto,
434 skb_vlan_tag_get(skb) & VLAN_VID_MASK);
436 return vlan ? vlan : br;
439 /* Some common code for IPv4/IPv6 */
440 struct net_device *setup_pre_routing(struct sk_buff *skb)
442 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
444 if (skb->pkt_type == PACKET_OTHERHOST) {
445 skb->pkt_type = PACKET_HOST;
446 nf_bridge->pkt_otherhost = true;
449 nf_bridge->in_prerouting = 1;
450 nf_bridge->physindev = skb->dev;
451 skb->dev = brnf_get_logical_dev(skb, skb->dev);
453 if (skb->protocol == htons(ETH_P_8021Q))
454 nf_bridge->orig_proto = BRNF_PROTO_8021Q;
455 else if (skb->protocol == htons(ETH_P_PPP_SES))
456 nf_bridge->orig_proto = BRNF_PROTO_PPPOE;
458 /* Must drop socket now because of tproxy. */
463 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
464 * Replicate the checks that IPv4 does on packet reception.
465 * Set skb->dev to the bridge device (i.e. parent of the
466 * receiving device) to make netfilter happy, the REDIRECT
467 * target in particular. Save the original destination IP
468 * address to be able to detect DNAT afterwards. */
469 static unsigned int br_nf_pre_routing(void *priv,
471 const struct nf_hook_state *state)
473 struct nf_bridge_info *nf_bridge;
474 struct net_bridge_port *p;
475 struct net_bridge *br;
476 __u32 len = nf_bridge_encap_header_len(skb);
478 if (unlikely(!pskb_may_pull(skb, len)))
481 p = br_port_get_rcu(state->in);
486 if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb)) {
487 if (!brnf_call_ip6tables && !br->nf_call_ip6tables)
490 nf_bridge_pull_encap_header_rcsum(skb);
491 return br_nf_pre_routing_ipv6(priv, skb, state);
494 if (!brnf_call_iptables && !br->nf_call_iptables)
497 if (!IS_IP(skb) && !IS_VLAN_IP(skb) && !IS_PPPOE_IP(skb))
500 nf_bridge_pull_encap_header_rcsum(skb);
502 if (br_validate_ipv4(state->net, skb))
505 nf_bridge_put(skb->nf_bridge);
506 if (!nf_bridge_alloc(skb))
508 if (!setup_pre_routing(skb))
511 nf_bridge = nf_bridge_info_get(skb);
512 nf_bridge->ipv4_daddr = ip_hdr(skb)->daddr;
514 skb->protocol = htons(ETH_P_IP);
515 skb->transport_header = skb->network_header + ip_hdr(skb)->ihl * 4;
517 NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, state->net, state->sk, skb,
519 br_nf_pre_routing_finish);
525 /* PF_BRIDGE/FORWARD *************************************************/
526 static int br_nf_forward_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
528 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
529 struct net_device *in;
531 if (!IS_ARP(skb) && !IS_VLAN_ARP(skb)) {
533 if (skb->protocol == htons(ETH_P_IP))
534 nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
536 if (skb->protocol == htons(ETH_P_IPV6))
537 nf_bridge->frag_max_size = IP6CB(skb)->frag_max_size;
539 in = nf_bridge->physindev;
540 if (nf_bridge->pkt_otherhost) {
541 skb->pkt_type = PACKET_OTHERHOST;
542 nf_bridge->pkt_otherhost = false;
544 nf_bridge_update_protocol(skb);
546 in = *((struct net_device **)(skb->cb));
548 nf_bridge_push_encap_header(skb);
550 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_FORWARD, net, sk, skb,
551 in, skb->dev, br_forward_finish, 1);
556 /* This is the 'purely bridged' case. For IP, we pass the packet to
557 * netfilter with indev and outdev set to the bridge device,
558 * but we are still able to filter on the 'real' indev/outdev
559 * because of the physdev module. For ARP, indev and outdev are the
561 static unsigned int br_nf_forward_ip(void *priv,
563 const struct nf_hook_state *state)
565 struct nf_bridge_info *nf_bridge;
566 struct net_device *parent;
572 /* Need exclusive nf_bridge_info since we might have multiple
573 * different physoutdevs. */
574 if (!nf_bridge_unshare(skb))
577 nf_bridge = nf_bridge_info_get(skb);
581 parent = bridge_parent(state->out);
585 if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
587 else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
592 nf_bridge_pull_encap_header(skb);
594 if (skb->pkt_type == PACKET_OTHERHOST) {
595 skb->pkt_type = PACKET_HOST;
596 nf_bridge->pkt_otherhost = true;
599 if (pf == NFPROTO_IPV4) {
600 if (br_validate_ipv4(state->net, skb))
602 IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
605 if (pf == NFPROTO_IPV6) {
606 if (br_validate_ipv6(state->net, skb))
608 IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
611 nf_bridge->physoutdev = skb->dev;
612 if (pf == NFPROTO_IPV4)
613 skb->protocol = htons(ETH_P_IP);
615 skb->protocol = htons(ETH_P_IPV6);
617 NF_HOOK(pf, NF_INET_FORWARD, state->net, NULL, skb,
618 brnf_get_logical_dev(skb, state->in),
619 parent, br_nf_forward_finish);
624 static unsigned int br_nf_forward_arp(void *priv,
626 const struct nf_hook_state *state)
628 struct net_bridge_port *p;
629 struct net_bridge *br;
630 struct net_device **d = (struct net_device **)(skb->cb);
632 p = br_port_get_rcu(state->out);
637 if (!brnf_call_arptables && !br->nf_call_arptables)
641 if (!IS_VLAN_ARP(skb))
643 nf_bridge_pull_encap_header(skb);
646 if (unlikely(!pskb_may_pull(skb, sizeof(struct arphdr))))
649 if (arp_hdr(skb)->ar_pln != 4) {
650 if (IS_VLAN_ARP(skb))
651 nf_bridge_push_encap_header(skb);
655 NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, state->net, state->sk, skb,
656 state->in, state->out, br_nf_forward_finish);
661 static int br_nf_push_frag_xmit(struct net *net, struct sock *sk, struct sk_buff *skb)
663 struct brnf_frag_data *data;
666 data = this_cpu_ptr(&brnf_frag_data_storage);
667 err = skb_cow_head(skb, data->size);
674 if (data->vlan_tci) {
675 skb->vlan_tci = data->vlan_tci;
676 skb->vlan_proto = data->vlan_proto;
679 skb_copy_to_linear_data_offset(skb, -data->size, data->mac, data->size);
680 __skb_push(skb, data->encap_size);
682 nf_bridge_info_free(skb);
683 return br_dev_queue_push_xmit(net, sk, skb);
687 br_nf_ip_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
688 int (*output)(struct net *, struct sock *, struct sk_buff *))
690 unsigned int mtu = ip_skb_dst_mtu(sk, skb);
691 struct iphdr *iph = ip_hdr(skb);
693 if (unlikely(((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) ||
694 (IPCB(skb)->frag_max_size &&
695 IPCB(skb)->frag_max_size > mtu))) {
696 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
701 return ip_do_fragment(net, sk, skb, output);
704 static unsigned int nf_bridge_mtu_reduction(const struct sk_buff *skb)
706 if (skb->nf_bridge->orig_proto == BRNF_PROTO_PPPOE)
707 return PPPOE_SES_HLEN;
711 static int br_nf_dev_queue_xmit(struct net *net, struct sock *sk, struct sk_buff *skb)
713 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
714 unsigned int mtu, mtu_reserved;
716 mtu_reserved = nf_bridge_mtu_reduction(skb);
719 if (nf_bridge->pkt_otherhost) {
720 skb->pkt_type = PACKET_OTHERHOST;
721 nf_bridge->pkt_otherhost = false;
724 if (nf_bridge->frag_max_size && nf_bridge->frag_max_size < mtu)
725 mtu = nf_bridge->frag_max_size;
727 nf_bridge_update_protocol(skb);
728 nf_bridge_push_encap_header(skb);
730 if (skb_is_gso(skb) || skb->len + mtu_reserved <= mtu) {
731 nf_bridge_info_free(skb);
732 return br_dev_queue_push_xmit(net, sk, skb);
735 /* This is wrong! We should preserve the original fragment
736 * boundaries by preserving frag_list rather than refragmenting.
738 if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV4) &&
739 skb->protocol == htons(ETH_P_IP)) {
740 struct brnf_frag_data *data;
742 if (br_validate_ipv4(net, skb))
745 IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
747 data = this_cpu_ptr(&brnf_frag_data_storage);
749 data->vlan_tci = skb->vlan_tci;
750 data->vlan_proto = skb->vlan_proto;
751 data->encap_size = nf_bridge_encap_header_len(skb);
752 data->size = ETH_HLEN + data->encap_size;
754 skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
757 return br_nf_ip_fragment(net, sk, skb, br_nf_push_frag_xmit);
759 if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV6) &&
760 skb->protocol == htons(ETH_P_IPV6)) {
761 const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops();
762 struct brnf_frag_data *data;
764 if (br_validate_ipv6(net, skb))
767 IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
769 data = this_cpu_ptr(&brnf_frag_data_storage);
770 data->encap_size = nf_bridge_encap_header_len(skb);
771 data->size = ETH_HLEN + data->encap_size;
773 skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
777 return v6ops->fragment(net, sk, skb, br_nf_push_frag_xmit);
782 nf_bridge_info_free(skb);
783 return br_dev_queue_push_xmit(net, sk, skb);
789 /* PF_BRIDGE/POST_ROUTING ********************************************/
790 static unsigned int br_nf_post_routing(void *priv,
792 const struct nf_hook_state *state)
794 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
795 struct net_device *realoutdev = bridge_parent(skb->dev);
798 /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
799 * on a bridge, but was delivered locally and is now being routed:
801 * POST_ROUTING was already invoked from the ip stack.
803 if (!nf_bridge || !nf_bridge->physoutdev)
809 if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
811 else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
816 if (skb->pkt_type == PACKET_OTHERHOST) {
817 skb->pkt_type = PACKET_HOST;
818 nf_bridge->pkt_otherhost = true;
821 nf_bridge_pull_encap_header(skb);
822 if (pf == NFPROTO_IPV4)
823 skb->protocol = htons(ETH_P_IP);
825 skb->protocol = htons(ETH_P_IPV6);
827 NF_HOOK(pf, NF_INET_POST_ROUTING, state->net, state->sk, skb,
829 br_nf_dev_queue_xmit);
834 /* IP/SABOTAGE *****************************************************/
835 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
836 * for the second time. */
837 static unsigned int ip_sabotage_in(void *priv,
839 const struct nf_hook_state *state)
841 if (skb->nf_bridge && !skb->nf_bridge->in_prerouting)
847 /* This is called when br_netfilter has called into iptables/netfilter,
848 * and DNAT has taken place on a bridge-forwarded packet.
850 * neigh->output has created a new MAC header, with local br0 MAC
853 * This restores the original MAC saddr of the bridged packet
854 * before invoking bridge forward logic to transmit the packet.
856 static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff *skb)
858 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
860 skb_pull(skb, ETH_HLEN);
861 nf_bridge->bridged_dnat = 0;
863 BUILD_BUG_ON(sizeof(nf_bridge->neigh_header) != (ETH_HLEN - ETH_ALEN));
865 skb_copy_to_linear_data_offset(skb, -(ETH_HLEN - ETH_ALEN),
866 nf_bridge->neigh_header,
867 ETH_HLEN - ETH_ALEN);
868 skb->dev = nf_bridge->physindev;
870 nf_bridge->physoutdev = NULL;
871 br_handle_frame_finish(dev_net(skb->dev), NULL, skb);
874 static int br_nf_dev_xmit(struct sk_buff *skb)
876 if (skb->nf_bridge && skb->nf_bridge->bridged_dnat) {
877 br_nf_pre_routing_finish_bridge_slow(skb);
883 static const struct nf_br_ops br_ops = {
884 .br_dev_xmit_hook = br_nf_dev_xmit,
887 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
888 * br_dev_queue_push_xmit is called afterwards */
889 static struct nf_hook_ops br_nf_ops[] __read_mostly = {
891 .hook = br_nf_pre_routing,
892 .pf = NFPROTO_BRIDGE,
893 .hooknum = NF_BR_PRE_ROUTING,
894 .priority = NF_BR_PRI_BRNF,
897 .hook = br_nf_forward_ip,
898 .pf = NFPROTO_BRIDGE,
899 .hooknum = NF_BR_FORWARD,
900 .priority = NF_BR_PRI_BRNF - 1,
903 .hook = br_nf_forward_arp,
904 .pf = NFPROTO_BRIDGE,
905 .hooknum = NF_BR_FORWARD,
906 .priority = NF_BR_PRI_BRNF,
909 .hook = br_nf_post_routing,
910 .pf = NFPROTO_BRIDGE,
911 .hooknum = NF_BR_POST_ROUTING,
912 .priority = NF_BR_PRI_LAST,
915 .hook = ip_sabotage_in,
917 .hooknum = NF_INET_PRE_ROUTING,
918 .priority = NF_IP_PRI_FIRST,
921 .hook = ip_sabotage_in,
923 .hooknum = NF_INET_PRE_ROUTING,
924 .priority = NF_IP6_PRI_FIRST,
928 static int brnf_device_event(struct notifier_block *unused, unsigned long event,
931 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
932 struct brnf_net *brnet;
936 if (event != NETDEV_REGISTER || !(dev->priv_flags & IFF_EBRIDGE))
942 brnet = net_generic(net, brnf_net_id);
946 ret = nf_register_net_hooks(net, br_nf_ops, ARRAY_SIZE(br_nf_ops));
950 brnet->enabled = true;
954 static void __net_exit brnf_exit_net(struct net *net)
956 struct brnf_net *brnet = net_generic(net, brnf_net_id);
961 nf_unregister_net_hooks(net, br_nf_ops, ARRAY_SIZE(br_nf_ops));
962 brnet->enabled = false;
965 static struct pernet_operations brnf_net_ops __read_mostly = {
966 .exit = brnf_exit_net,
968 .size = sizeof(struct brnf_net),
971 static struct notifier_block brnf_notifier __read_mostly = {
972 .notifier_call = brnf_device_event,
975 /* recursively invokes nf_hook_slow (again), skipping already-called
976 * hooks (< NF_BR_PRI_BRNF).
978 * Called with rcu read lock held.
980 int br_nf_hook_thresh(unsigned int hook, struct net *net,
981 struct sock *sk, struct sk_buff *skb,
982 struct net_device *indev,
983 struct net_device *outdev,
984 int (*okfn)(struct net *, struct sock *,
987 struct nf_hook_entry *elem;
988 struct nf_hook_state state;
991 elem = rcu_dereference(net->nf.hooks[NFPROTO_BRIDGE][hook]);
993 while (elem && (elem->ops.priority <= NF_BR_PRI_BRNF))
994 elem = rcu_dereference(elem->next);
997 return okfn(net, sk, skb);
999 /* We may already have this, but read-locks nest anyway */
1001 nf_hook_state_init(&state, elem, hook, NF_BR_PRI_BRNF + 1,
1002 NFPROTO_BRIDGE, indev, outdev, sk, net, okfn);
1004 ret = nf_hook_slow(skb, &state);
1007 ret = okfn(net, sk, skb);
1012 #ifdef CONFIG_SYSCTL
1014 int brnf_sysctl_call_tables(struct ctl_table *ctl, int write,
1015 void __user *buffer, size_t *lenp, loff_t *ppos)
1019 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1021 if (write && *(int *)(ctl->data))
1022 *(int *)(ctl->data) = 1;
1026 static struct ctl_table brnf_table[] = {
1028 .procname = "bridge-nf-call-arptables",
1029 .data = &brnf_call_arptables,
1030 .maxlen = sizeof(int),
1032 .proc_handler = brnf_sysctl_call_tables,
1035 .procname = "bridge-nf-call-iptables",
1036 .data = &brnf_call_iptables,
1037 .maxlen = sizeof(int),
1039 .proc_handler = brnf_sysctl_call_tables,
1042 .procname = "bridge-nf-call-ip6tables",
1043 .data = &brnf_call_ip6tables,
1044 .maxlen = sizeof(int),
1046 .proc_handler = brnf_sysctl_call_tables,
1049 .procname = "bridge-nf-filter-vlan-tagged",
1050 .data = &brnf_filter_vlan_tagged,
1051 .maxlen = sizeof(int),
1053 .proc_handler = brnf_sysctl_call_tables,
1056 .procname = "bridge-nf-filter-pppoe-tagged",
1057 .data = &brnf_filter_pppoe_tagged,
1058 .maxlen = sizeof(int),
1060 .proc_handler = brnf_sysctl_call_tables,
1063 .procname = "bridge-nf-pass-vlan-input-dev",
1064 .data = &brnf_pass_vlan_indev,
1065 .maxlen = sizeof(int),
1067 .proc_handler = brnf_sysctl_call_tables,
1073 static int __init br_netfilter_init(void)
1077 ret = register_pernet_subsys(&brnf_net_ops);
1081 ret = register_netdevice_notifier(&brnf_notifier);
1083 unregister_pernet_subsys(&brnf_net_ops);
1087 #ifdef CONFIG_SYSCTL
1088 brnf_sysctl_header = register_net_sysctl(&init_net, "net/bridge", brnf_table);
1089 if (brnf_sysctl_header == NULL) {
1091 "br_netfilter: can't register to sysctl.\n");
1092 unregister_netdevice_notifier(&brnf_notifier);
1093 unregister_pernet_subsys(&brnf_net_ops);
1097 RCU_INIT_POINTER(nf_br_ops, &br_ops);
1098 printk(KERN_NOTICE "Bridge firewalling registered\n");
1102 static void __exit br_netfilter_fini(void)
1104 RCU_INIT_POINTER(nf_br_ops, NULL);
1105 unregister_netdevice_notifier(&brnf_notifier);
1106 unregister_pernet_subsys(&brnf_net_ops);
1107 #ifdef CONFIG_SYSCTL
1108 unregister_net_sysctl_table(brnf_sysctl_header);
1112 module_init(br_netfilter_init);
1113 module_exit(br_netfilter_fini);
1115 MODULE_LICENSE("GPL");
1116 MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
1117 MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
1118 MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");