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 <linux/uaccess.h>
44 #include "br_private.h"
46 #include <linux/sysctl.h>
49 static unsigned 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 (refcount_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 refcount_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_CSUMERRORS);
242 __IP_INC_STATS(net, IPSTATS_MIB_INHDRERRORS);
247 void nf_bridge_update_protocol(struct sk_buff *skb)
249 switch (skb->nf_bridge->orig_proto) {
250 case BRNF_PROTO_8021Q:
251 skb->protocol = htons(ETH_P_8021Q);
253 case BRNF_PROTO_PPPOE:
254 skb->protocol = htons(ETH_P_PPP_SES);
256 case BRNF_PROTO_UNCHANGED:
261 /* Obtain the correct destination MAC address, while preserving the original
262 * source MAC address. If we already know this address, we just copy it. If we
263 * don't, we use the neighbour framework to find out. In both cases, we make
264 * sure that br_handle_frame_finish() is called afterwards.
266 int br_nf_pre_routing_finish_bridge(struct net *net, struct sock *sk, struct sk_buff *skb)
268 struct neighbour *neigh;
269 struct dst_entry *dst;
271 skb->dev = bridge_parent(skb->dev);
275 neigh = dst_neigh_lookup_skb(dst, skb);
277 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
280 if (neigh->hh.hh_len) {
281 neigh_hh_bridge(&neigh->hh, skb);
282 skb->dev = nf_bridge->physindev;
283 ret = br_handle_frame_finish(net, sk, skb);
285 /* the neighbour function below overwrites the complete
286 * MAC header, so we save the Ethernet source address and
289 skb_copy_from_linear_data_offset(skb,
290 -(ETH_HLEN-ETH_ALEN),
291 nf_bridge->neigh_header,
293 /* tell br_dev_xmit to continue with forwarding */
294 nf_bridge->bridged_dnat = 1;
295 /* FIXME Need to refragment */
296 ret = neigh->output(neigh, skb);
298 neigh_release(neigh);
307 br_nf_ipv4_daddr_was_changed(const struct sk_buff *skb,
308 const struct nf_bridge_info *nf_bridge)
310 return ip_hdr(skb)->daddr != nf_bridge->ipv4_daddr;
313 /* This requires some explaining. If DNAT has taken place,
314 * we will need to fix up the destination Ethernet address.
315 * This is also true when SNAT takes place (for the reply direction).
317 * There are two cases to consider:
318 * 1. The packet was DNAT'ed to a device in the same bridge
319 * port group as it was received on. We can still bridge
321 * 2. The packet was DNAT'ed to a different device, either
322 * a non-bridged device or another bridge port group.
323 * The packet will need to be routed.
325 * The correct way of distinguishing between these two cases is to
326 * call ip_route_input() and to look at skb->dst->dev, which is
327 * changed to the destination device if ip_route_input() succeeds.
329 * Let's first consider the case that ip_route_input() succeeds:
331 * If the output device equals the logical bridge device the packet
332 * came in on, we can consider this bridging. The corresponding MAC
333 * address will be obtained in br_nf_pre_routing_finish_bridge.
334 * Otherwise, the packet is considered to be routed and we just
335 * change the destination MAC address so that the packet will
336 * later be passed up to the IP stack to be routed. For a redirected
337 * packet, ip_route_input() will give back the localhost as output device,
338 * which differs from the bridge device.
340 * Let's now consider the case that ip_route_input() fails:
342 * This can be because the destination address is martian, in which case
343 * the packet will be dropped.
344 * If IP forwarding is disabled, ip_route_input() will fail, while
345 * ip_route_output_key() can return success. The source
346 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
347 * thinks we're handling a locally generated packet and won't care
348 * if IP forwarding is enabled. If the output device equals the logical bridge
349 * device, we proceed as if ip_route_input() succeeded. If it differs from the
350 * logical bridge port or if ip_route_output_key() fails we drop the packet.
352 static int br_nf_pre_routing_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
354 struct net_device *dev = skb->dev;
355 struct iphdr *iph = ip_hdr(skb);
356 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
360 nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
362 if (nf_bridge->pkt_otherhost) {
363 skb->pkt_type = PACKET_OTHERHOST;
364 nf_bridge->pkt_otherhost = false;
366 nf_bridge->in_prerouting = 0;
367 if (br_nf_ipv4_daddr_was_changed(skb, nf_bridge)) {
368 if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
369 struct in_device *in_dev = __in_dev_get_rcu(dev);
371 /* If err equals -EHOSTUNREACH the error is due to a
372 * martian destination or due to the fact that
373 * forwarding is disabled. For most martian packets,
374 * ip_route_output_key() will fail. It won't fail for 2 types of
375 * martian destinations: loopback destinations and destination
376 * 0.0.0.0. In both cases the packet will be dropped because the
377 * destination is the loopback device and not the bridge. */
378 if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
381 rt = ip_route_output(net, iph->daddr, 0,
382 RT_TOS(iph->tos), 0);
384 /* - Bridged-and-DNAT'ed traffic doesn't
385 * require ip_forwarding. */
386 if (rt->dst.dev == dev) {
387 skb_dst_set(skb, &rt->dst);
396 if (skb_dst(skb)->dev == dev) {
398 skb->dev = nf_bridge->physindev;
399 nf_bridge_update_protocol(skb);
400 nf_bridge_push_encap_header(skb);
401 br_nf_hook_thresh(NF_BR_PRE_ROUTING,
402 net, sk, skb, skb->dev,
404 br_nf_pre_routing_finish_bridge);
407 ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr);
408 skb->pkt_type = PACKET_HOST;
411 rt = bridge_parent_rtable(nf_bridge->physindev);
416 skb_dst_set_noref(skb, &rt->dst);
419 skb->dev = nf_bridge->physindev;
420 nf_bridge_update_protocol(skb);
421 nf_bridge_push_encap_header(skb);
422 br_nf_hook_thresh(NF_BR_PRE_ROUTING, net, sk, skb, skb->dev, NULL,
423 br_handle_frame_finish);
427 static struct net_device *brnf_get_logical_dev(struct sk_buff *skb, const struct net_device *dev)
429 struct net_device *vlan, *br;
431 br = bridge_parent(dev);
432 if (brnf_pass_vlan_indev == 0 || !skb_vlan_tag_present(skb))
435 vlan = __vlan_find_dev_deep_rcu(br, skb->vlan_proto,
436 skb_vlan_tag_get(skb) & VLAN_VID_MASK);
438 return vlan ? vlan : br;
441 /* Some common code for IPv4/IPv6 */
442 struct net_device *setup_pre_routing(struct sk_buff *skb)
444 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
446 if (skb->pkt_type == PACKET_OTHERHOST) {
447 skb->pkt_type = PACKET_HOST;
448 nf_bridge->pkt_otherhost = true;
451 nf_bridge->in_prerouting = 1;
452 nf_bridge->physindev = skb->dev;
453 skb->dev = brnf_get_logical_dev(skb, skb->dev);
455 if (skb->protocol == htons(ETH_P_8021Q))
456 nf_bridge->orig_proto = BRNF_PROTO_8021Q;
457 else if (skb->protocol == htons(ETH_P_PPP_SES))
458 nf_bridge->orig_proto = BRNF_PROTO_PPPOE;
460 /* Must drop socket now because of tproxy. */
465 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
466 * Replicate the checks that IPv4 does on packet reception.
467 * Set skb->dev to the bridge device (i.e. parent of the
468 * receiving device) to make netfilter happy, the REDIRECT
469 * target in particular. Save the original destination IP
470 * address to be able to detect DNAT afterwards. */
471 static unsigned int br_nf_pre_routing(void *priv,
473 const struct nf_hook_state *state)
475 struct nf_bridge_info *nf_bridge;
476 struct net_bridge_port *p;
477 struct net_bridge *br;
478 __u32 len = nf_bridge_encap_header_len(skb);
480 if (unlikely(!pskb_may_pull(skb, len)))
483 p = br_port_get_rcu(state->in);
488 if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb)) {
489 if (!brnf_call_ip6tables && !br->nf_call_ip6tables)
492 nf_bridge_pull_encap_header_rcsum(skb);
493 return br_nf_pre_routing_ipv6(priv, skb, state);
496 if (!brnf_call_iptables && !br->nf_call_iptables)
499 if (!IS_IP(skb) && !IS_VLAN_IP(skb) && !IS_PPPOE_IP(skb))
502 nf_bridge_pull_encap_header_rcsum(skb);
504 if (br_validate_ipv4(state->net, skb))
507 nf_bridge_put(skb->nf_bridge);
508 if (!nf_bridge_alloc(skb))
510 if (!setup_pre_routing(skb))
513 nf_bridge = nf_bridge_info_get(skb);
514 nf_bridge->ipv4_daddr = ip_hdr(skb)->daddr;
516 skb->protocol = htons(ETH_P_IP);
518 NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, state->net, state->sk, skb,
520 br_nf_pre_routing_finish);
526 /* PF_BRIDGE/FORWARD *************************************************/
527 static int br_nf_forward_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
529 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
530 struct net_device *in;
532 if (!IS_ARP(skb) && !IS_VLAN_ARP(skb)) {
534 if (skb->protocol == htons(ETH_P_IP))
535 nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
537 if (skb->protocol == htons(ETH_P_IPV6))
538 nf_bridge->frag_max_size = IP6CB(skb)->frag_max_size;
540 in = nf_bridge->physindev;
541 if (nf_bridge->pkt_otherhost) {
542 skb->pkt_type = PACKET_OTHERHOST;
543 nf_bridge->pkt_otherhost = false;
545 nf_bridge_update_protocol(skb);
547 in = *((struct net_device **)(skb->cb));
549 nf_bridge_push_encap_header(skb);
551 br_nf_hook_thresh(NF_BR_FORWARD, net, sk, skb, in, skb->dev,
557 /* This is the 'purely bridged' case. For IP, we pass the packet to
558 * netfilter with indev and outdev set to the bridge device,
559 * but we are still able to filter on the 'real' indev/outdev
560 * because of the physdev module. For ARP, indev and outdev are the
562 static unsigned int br_nf_forward_ip(void *priv,
564 const struct nf_hook_state *state)
566 struct nf_bridge_info *nf_bridge;
567 struct net_device *parent;
573 /* Need exclusive nf_bridge_info since we might have multiple
574 * different physoutdevs. */
575 if (!nf_bridge_unshare(skb))
578 nf_bridge = nf_bridge_info_get(skb);
582 parent = bridge_parent(state->out);
586 if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
588 else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
593 nf_bridge_pull_encap_header(skb);
595 if (skb->pkt_type == PACKET_OTHERHOST) {
596 skb->pkt_type = PACKET_HOST;
597 nf_bridge->pkt_otherhost = true;
600 if (pf == NFPROTO_IPV4) {
601 if (br_validate_ipv4(state->net, skb))
603 IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
606 if (pf == NFPROTO_IPV6) {
607 if (br_validate_ipv6(state->net, skb))
609 IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
612 nf_bridge->physoutdev = skb->dev;
613 if (pf == NFPROTO_IPV4)
614 skb->protocol = htons(ETH_P_IP);
616 skb->protocol = htons(ETH_P_IPV6);
618 NF_HOOK(pf, NF_INET_FORWARD, state->net, NULL, skb,
619 brnf_get_logical_dev(skb, state->in),
620 parent, br_nf_forward_finish);
625 static unsigned int br_nf_forward_arp(void *priv,
627 const struct nf_hook_state *state)
629 struct net_bridge_port *p;
630 struct net_bridge *br;
631 struct net_device **d = (struct net_device **)(skb->cb);
633 p = br_port_get_rcu(state->out);
638 if (!brnf_call_arptables && !br->nf_call_arptables)
642 if (!IS_VLAN_ARP(skb))
644 nf_bridge_pull_encap_header(skb);
647 if (arp_hdr(skb)->ar_pln != 4) {
648 if (IS_VLAN_ARP(skb))
649 nf_bridge_push_encap_header(skb);
653 NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, state->net, state->sk, skb,
654 state->in, state->out, br_nf_forward_finish);
659 static int br_nf_push_frag_xmit(struct net *net, struct sock *sk, struct sk_buff *skb)
661 struct brnf_frag_data *data;
664 data = this_cpu_ptr(&brnf_frag_data_storage);
665 err = skb_cow_head(skb, data->size);
672 if (data->vlan_tci) {
673 skb->vlan_tci = data->vlan_tci;
674 skb->vlan_proto = data->vlan_proto;
677 skb_copy_to_linear_data_offset(skb, -data->size, data->mac, data->size);
678 __skb_push(skb, data->encap_size);
680 nf_bridge_info_free(skb);
681 return br_dev_queue_push_xmit(net, sk, skb);
685 br_nf_ip_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
686 int (*output)(struct net *, struct sock *, struct sk_buff *))
688 unsigned int mtu = ip_skb_dst_mtu(sk, skb);
689 struct iphdr *iph = ip_hdr(skb);
691 if (unlikely(((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) ||
692 (IPCB(skb)->frag_max_size &&
693 IPCB(skb)->frag_max_size > mtu))) {
694 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
699 return ip_do_fragment(net, sk, skb, output);
702 static unsigned int nf_bridge_mtu_reduction(const struct sk_buff *skb)
704 if (skb->nf_bridge->orig_proto == BRNF_PROTO_PPPOE)
705 return PPPOE_SES_HLEN;
709 static int br_nf_dev_queue_xmit(struct net *net, struct sock *sk, struct sk_buff *skb)
711 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
712 unsigned int mtu, mtu_reserved;
714 mtu_reserved = nf_bridge_mtu_reduction(skb);
717 if (nf_bridge->frag_max_size && nf_bridge->frag_max_size < mtu)
718 mtu = nf_bridge->frag_max_size;
720 if (skb_is_gso(skb) || skb->len + mtu_reserved <= mtu) {
721 nf_bridge_info_free(skb);
722 return br_dev_queue_push_xmit(net, sk, skb);
725 /* This is wrong! We should preserve the original fragment
726 * boundaries by preserving frag_list rather than refragmenting.
728 if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV4) &&
729 skb->protocol == htons(ETH_P_IP)) {
730 struct brnf_frag_data *data;
732 if (br_validate_ipv4(net, skb))
735 IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
737 nf_bridge_update_protocol(skb);
739 data = this_cpu_ptr(&brnf_frag_data_storage);
741 data->vlan_tci = skb->vlan_tci;
742 data->vlan_proto = skb->vlan_proto;
743 data->encap_size = nf_bridge_encap_header_len(skb);
744 data->size = ETH_HLEN + data->encap_size;
746 skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
749 return br_nf_ip_fragment(net, sk, skb, br_nf_push_frag_xmit);
751 if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV6) &&
752 skb->protocol == htons(ETH_P_IPV6)) {
753 const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops();
754 struct brnf_frag_data *data;
756 if (br_validate_ipv6(net, skb))
759 IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
761 nf_bridge_update_protocol(skb);
763 data = this_cpu_ptr(&brnf_frag_data_storage);
764 data->encap_size = nf_bridge_encap_header_len(skb);
765 data->size = ETH_HLEN + data->encap_size;
767 skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
771 return v6ops->fragment(net, sk, skb, br_nf_push_frag_xmit);
776 nf_bridge_info_free(skb);
777 return br_dev_queue_push_xmit(net, sk, skb);
783 /* PF_BRIDGE/POST_ROUTING ********************************************/
784 static unsigned int br_nf_post_routing(void *priv,
786 const struct nf_hook_state *state)
788 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
789 struct net_device *realoutdev = bridge_parent(skb->dev);
792 /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
793 * on a bridge, but was delivered locally and is now being routed:
795 * POST_ROUTING was already invoked from the ip stack.
797 if (!nf_bridge || !nf_bridge->physoutdev)
803 if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
805 else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
810 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
811 * about the value of skb->pkt_type. */
812 if (skb->pkt_type == PACKET_OTHERHOST) {
813 skb->pkt_type = PACKET_HOST;
814 nf_bridge->pkt_otherhost = true;
817 nf_bridge_pull_encap_header(skb);
818 if (pf == NFPROTO_IPV4)
819 skb->protocol = htons(ETH_P_IP);
821 skb->protocol = htons(ETH_P_IPV6);
823 NF_HOOK(pf, NF_INET_POST_ROUTING, state->net, state->sk, skb,
825 br_nf_dev_queue_xmit);
830 /* IP/SABOTAGE *****************************************************/
831 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
832 * for the second time. */
833 static unsigned int ip_sabotage_in(void *priv,
835 const struct nf_hook_state *state)
837 if (skb->nf_bridge && !skb->nf_bridge->in_prerouting) {
838 state->okfn(state->net, state->sk, skb);
845 /* This is called when br_netfilter has called into iptables/netfilter,
846 * and DNAT has taken place on a bridge-forwarded packet.
848 * neigh->output has created a new MAC header, with local br0 MAC
851 * This restores the original MAC saddr of the bridged packet
852 * before invoking bridge forward logic to transmit the packet.
854 static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff *skb)
856 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
858 skb_pull(skb, ETH_HLEN);
859 nf_bridge->bridged_dnat = 0;
861 BUILD_BUG_ON(sizeof(nf_bridge->neigh_header) != (ETH_HLEN - ETH_ALEN));
863 skb_copy_to_linear_data_offset(skb, -(ETH_HLEN - ETH_ALEN),
864 nf_bridge->neigh_header,
865 ETH_HLEN - ETH_ALEN);
866 skb->dev = nf_bridge->physindev;
868 nf_bridge->physoutdev = NULL;
869 br_handle_frame_finish(dev_net(skb->dev), NULL, skb);
872 static int br_nf_dev_xmit(struct sk_buff *skb)
874 if (skb->nf_bridge && skb->nf_bridge->bridged_dnat) {
875 br_nf_pre_routing_finish_bridge_slow(skb);
881 static const struct nf_br_ops br_ops = {
882 .br_dev_xmit_hook = br_nf_dev_xmit,
885 void br_netfilter_enable(void)
888 EXPORT_SYMBOL_GPL(br_netfilter_enable);
890 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
891 * br_dev_queue_push_xmit is called afterwards */
892 static const struct nf_hook_ops br_nf_ops[] = {
894 .hook = br_nf_pre_routing,
895 .pf = NFPROTO_BRIDGE,
896 .hooknum = NF_BR_PRE_ROUTING,
897 .priority = NF_BR_PRI_BRNF,
900 .hook = br_nf_forward_ip,
901 .pf = NFPROTO_BRIDGE,
902 .hooknum = NF_BR_FORWARD,
903 .priority = NF_BR_PRI_BRNF - 1,
906 .hook = br_nf_forward_arp,
907 .pf = NFPROTO_BRIDGE,
908 .hooknum = NF_BR_FORWARD,
909 .priority = NF_BR_PRI_BRNF,
912 .hook = br_nf_post_routing,
913 .pf = NFPROTO_BRIDGE,
914 .hooknum = NF_BR_POST_ROUTING,
915 .priority = NF_BR_PRI_LAST,
918 .hook = ip_sabotage_in,
920 .hooknum = NF_INET_PRE_ROUTING,
921 .priority = NF_IP_PRI_FIRST,
924 .hook = ip_sabotage_in,
926 .hooknum = NF_INET_PRE_ROUTING,
927 .priority = NF_IP6_PRI_FIRST,
931 static int brnf_device_event(struct notifier_block *unused, unsigned long event,
934 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
935 struct brnf_net *brnet;
939 if (event != NETDEV_REGISTER || !(dev->priv_flags & IFF_EBRIDGE))
945 brnet = net_generic(net, brnf_net_id);
949 ret = nf_register_net_hooks(net, br_nf_ops, ARRAY_SIZE(br_nf_ops));
953 brnet->enabled = true;
957 static void __net_exit brnf_exit_net(struct net *net)
959 struct brnf_net *brnet = net_generic(net, brnf_net_id);
964 nf_unregister_net_hooks(net, br_nf_ops, ARRAY_SIZE(br_nf_ops));
965 brnet->enabled = false;
968 static struct pernet_operations brnf_net_ops __read_mostly = {
969 .exit = brnf_exit_net,
971 .size = sizeof(struct brnf_net),
974 static struct notifier_block brnf_notifier __read_mostly = {
975 .notifier_call = brnf_device_event,
978 /* recursively invokes nf_hook_slow (again), skipping already-called
979 * hooks (< NF_BR_PRI_BRNF).
981 * Called with rcu read lock held.
983 int br_nf_hook_thresh(unsigned int hook, struct net *net,
984 struct sock *sk, struct sk_buff *skb,
985 struct net_device *indev,
986 struct net_device *outdev,
987 int (*okfn)(struct net *, struct sock *,
990 const struct nf_hook_entries *e;
991 struct nf_hook_state state;
992 struct nf_hook_ops **ops;
996 e = rcu_dereference(net->nf.hooks_bridge[hook]);
998 return okfn(net, sk, skb);
1000 ops = nf_hook_entries_get_hook_ops(e);
1001 for (i = 0; i < e->num_hook_entries &&
1002 ops[i]->priority <= NF_BR_PRI_BRNF; i++)
1005 nf_hook_state_init(&state, hook, NFPROTO_BRIDGE, indev, outdev,
1008 ret = nf_hook_slow(skb, &state, e, i);
1010 ret = okfn(net, sk, skb);
1015 #ifdef CONFIG_SYSCTL
1017 int brnf_sysctl_call_tables(struct ctl_table *ctl, int write,
1018 void __user *buffer, size_t *lenp, loff_t *ppos)
1022 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1024 if (write && *(int *)(ctl->data))
1025 *(int *)(ctl->data) = 1;
1029 static struct ctl_table brnf_table[] = {
1031 .procname = "bridge-nf-call-arptables",
1032 .data = &brnf_call_arptables,
1033 .maxlen = sizeof(int),
1035 .proc_handler = brnf_sysctl_call_tables,
1038 .procname = "bridge-nf-call-iptables",
1039 .data = &brnf_call_iptables,
1040 .maxlen = sizeof(int),
1042 .proc_handler = brnf_sysctl_call_tables,
1045 .procname = "bridge-nf-call-ip6tables",
1046 .data = &brnf_call_ip6tables,
1047 .maxlen = sizeof(int),
1049 .proc_handler = brnf_sysctl_call_tables,
1052 .procname = "bridge-nf-filter-vlan-tagged",
1053 .data = &brnf_filter_vlan_tagged,
1054 .maxlen = sizeof(int),
1056 .proc_handler = brnf_sysctl_call_tables,
1059 .procname = "bridge-nf-filter-pppoe-tagged",
1060 .data = &brnf_filter_pppoe_tagged,
1061 .maxlen = sizeof(int),
1063 .proc_handler = brnf_sysctl_call_tables,
1066 .procname = "bridge-nf-pass-vlan-input-dev",
1067 .data = &brnf_pass_vlan_indev,
1068 .maxlen = sizeof(int),
1070 .proc_handler = brnf_sysctl_call_tables,
1076 static int __init br_netfilter_init(void)
1080 ret = register_pernet_subsys(&brnf_net_ops);
1084 ret = register_netdevice_notifier(&brnf_notifier);
1086 unregister_pernet_subsys(&brnf_net_ops);
1090 #ifdef CONFIG_SYSCTL
1091 brnf_sysctl_header = register_net_sysctl(&init_net, "net/bridge", brnf_table);
1092 if (brnf_sysctl_header == NULL) {
1094 "br_netfilter: can't register to sysctl.\n");
1095 unregister_netdevice_notifier(&brnf_notifier);
1096 unregister_pernet_subsys(&brnf_net_ops);
1100 RCU_INIT_POINTER(nf_br_ops, &br_ops);
1101 printk(KERN_NOTICE "Bridge firewalling registered\n");
1105 static void __exit br_netfilter_fini(void)
1107 RCU_INIT_POINTER(nf_br_ops, NULL);
1108 unregister_netdevice_notifier(&brnf_notifier);
1109 unregister_pernet_subsys(&brnf_net_ops);
1110 #ifdef CONFIG_SYSCTL
1111 unregister_net_sysctl_table(brnf_sysctl_header);
1115 module_init(br_netfilter_init);
1116 module_exit(br_netfilter_fini);
1118 MODULE_LICENSE("GPL");
1119 MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
1120 MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
1121 MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");