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 <uapi/linux/netfilter_bridge.h>
30 #include <linux/netfilter_ipv4.h>
31 #include <linux/netfilter_ipv6.h>
32 #include <linux/netfilter_arp.h>
33 #include <linux/in_route.h>
34 #include <linux/rculist.h>
35 #include <linux/inetdevice.h>
39 #include <net/addrconf.h>
40 #include <net/route.h>
41 #include <net/netfilter/br_netfilter.h>
42 #include <net/netns/generic.h>
44 #include <linux/uaccess.h>
45 #include "br_private.h"
47 #include <linux/sysctl.h>
50 static unsigned int brnf_net_id __read_mostly;
57 static struct ctl_table_header *brnf_sysctl_header;
58 static int brnf_call_iptables __read_mostly = 1;
59 static int brnf_call_ip6tables __read_mostly = 1;
60 static int brnf_call_arptables __read_mostly = 1;
61 static int brnf_filter_vlan_tagged __read_mostly;
62 static int brnf_filter_pppoe_tagged __read_mostly;
63 static int brnf_pass_vlan_indev __read_mostly;
65 #define brnf_call_iptables 1
66 #define brnf_call_ip6tables 1
67 #define brnf_call_arptables 1
68 #define brnf_filter_vlan_tagged 0
69 #define brnf_filter_pppoe_tagged 0
70 #define brnf_pass_vlan_indev 0
74 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP))
76 #define IS_IPV6(skb) \
77 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6))
80 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP))
82 static inline __be16 vlan_proto(const struct sk_buff *skb)
84 if (skb_vlan_tag_present(skb))
86 else if (skb->protocol == htons(ETH_P_8021Q))
87 return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
92 #define IS_VLAN_IP(skb) \
93 (vlan_proto(skb) == htons(ETH_P_IP) && \
94 brnf_filter_vlan_tagged)
96 #define IS_VLAN_IPV6(skb) \
97 (vlan_proto(skb) == htons(ETH_P_IPV6) && \
98 brnf_filter_vlan_tagged)
100 #define IS_VLAN_ARP(skb) \
101 (vlan_proto(skb) == htons(ETH_P_ARP) && \
102 brnf_filter_vlan_tagged)
104 static inline __be16 pppoe_proto(const struct sk_buff *skb)
106 return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
107 sizeof(struct pppoe_hdr)));
110 #define IS_PPPOE_IP(skb) \
111 (skb->protocol == htons(ETH_P_PPP_SES) && \
112 pppoe_proto(skb) == htons(PPP_IP) && \
113 brnf_filter_pppoe_tagged)
115 #define IS_PPPOE_IPV6(skb) \
116 (skb->protocol == htons(ETH_P_PPP_SES) && \
117 pppoe_proto(skb) == htons(PPP_IPV6) && \
118 brnf_filter_pppoe_tagged)
120 /* largest possible L2 header, see br_nf_dev_queue_xmit() */
121 #define NF_BRIDGE_MAX_MAC_HEADER_LENGTH (PPPOE_SES_HLEN + ETH_HLEN)
123 struct brnf_frag_data {
124 char mac[NF_BRIDGE_MAX_MAC_HEADER_LENGTH];
131 static DEFINE_PER_CPU(struct brnf_frag_data, brnf_frag_data_storage);
133 static void nf_bridge_info_free(struct sk_buff *skb)
135 if (skb->nf_bridge) {
136 nf_bridge_put(skb->nf_bridge);
137 skb->nf_bridge = NULL;
141 static inline struct net_device *bridge_parent(const struct net_device *dev)
143 struct net_bridge_port *port;
145 port = br_port_get_rcu(dev);
146 return port ? port->br->dev : NULL;
149 static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
151 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
153 if (refcount_read(&nf_bridge->use) > 1) {
154 struct nf_bridge_info *tmp = nf_bridge_alloc(skb);
157 memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info));
158 refcount_set(&tmp->use, 1);
160 nf_bridge_put(nf_bridge);
166 unsigned int nf_bridge_encap_header_len(const struct sk_buff *skb)
168 switch (skb->protocol) {
169 case __cpu_to_be16(ETH_P_8021Q):
171 case __cpu_to_be16(ETH_P_PPP_SES):
172 return PPPOE_SES_HLEN;
178 static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
180 unsigned int len = nf_bridge_encap_header_len(skb);
183 skb->network_header += len;
186 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
188 unsigned int len = nf_bridge_encap_header_len(skb);
190 skb_pull_rcsum(skb, len);
191 skb->network_header += len;
194 /* When handing a packet over to the IP layer
195 * check whether we have a skb that is in the
199 static int br_validate_ipv4(struct net *net, struct sk_buff *skb)
201 const struct iphdr *iph;
204 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
209 /* Basic sanity checks */
210 if (iph->ihl < 5 || iph->version != 4)
213 if (!pskb_may_pull(skb, iph->ihl*4))
217 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
220 len = ntohs(iph->tot_len);
221 if (skb->len < len) {
222 __IP_INC_STATS(net, IPSTATS_MIB_INTRUNCATEDPKTS);
224 } else if (len < (iph->ihl*4))
227 if (pskb_trim_rcsum(skb, len)) {
228 __IP_INC_STATS(net, IPSTATS_MIB_INDISCARDS);
232 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
233 /* We should really parse IP options here but until
234 * somebody who actually uses IP options complains to
235 * us we'll just silently ignore the options because
241 __IP_INC_STATS(net, IPSTATS_MIB_CSUMERRORS);
243 __IP_INC_STATS(net, IPSTATS_MIB_INHDRERRORS);
248 void nf_bridge_update_protocol(struct sk_buff *skb)
250 switch (skb->nf_bridge->orig_proto) {
251 case BRNF_PROTO_8021Q:
252 skb->protocol = htons(ETH_P_8021Q);
254 case BRNF_PROTO_PPPOE:
255 skb->protocol = htons(ETH_P_PPP_SES);
257 case BRNF_PROTO_UNCHANGED:
262 /* Obtain the correct destination MAC address, while preserving the original
263 * source MAC address. If we already know this address, we just copy it. If we
264 * don't, we use the neighbour framework to find out. In both cases, we make
265 * sure that br_handle_frame_finish() is called afterwards.
267 int br_nf_pre_routing_finish_bridge(struct net *net, struct sock *sk, struct sk_buff *skb)
269 struct neighbour *neigh;
270 struct dst_entry *dst;
272 skb->dev = bridge_parent(skb->dev);
276 neigh = dst_neigh_lookup_skb(dst, skb);
278 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
281 if ((neigh->nud_state & NUD_CONNECTED) && neigh->hh.hh_len) {
282 neigh_hh_bridge(&neigh->hh, skb);
283 skb->dev = nf_bridge->physindev;
284 ret = br_handle_frame_finish(net, sk, skb);
286 /* the neighbour function below overwrites the complete
287 * MAC header, so we save the Ethernet source address and
290 skb_copy_from_linear_data_offset(skb,
291 -(ETH_HLEN-ETH_ALEN),
292 nf_bridge->neigh_header,
294 /* tell br_dev_xmit to continue with forwarding */
295 nf_bridge->bridged_dnat = 1;
296 /* FIXME Need to refragment */
297 ret = neigh->output(neigh, skb);
299 neigh_release(neigh);
308 br_nf_ipv4_daddr_was_changed(const struct sk_buff *skb,
309 const struct nf_bridge_info *nf_bridge)
311 return ip_hdr(skb)->daddr != nf_bridge->ipv4_daddr;
314 /* This requires some explaining. If DNAT has taken place,
315 * we will need to fix up the destination Ethernet address.
316 * This is also true when SNAT takes place (for the reply direction).
318 * There are two cases to consider:
319 * 1. The packet was DNAT'ed to a device in the same bridge
320 * port group as it was received on. We can still bridge
322 * 2. The packet was DNAT'ed to a different device, either
323 * a non-bridged device or another bridge port group.
324 * The packet will need to be routed.
326 * The correct way of distinguishing between these two cases is to
327 * call ip_route_input() and to look at skb->dst->dev, which is
328 * changed to the destination device if ip_route_input() succeeds.
330 * Let's first consider the case that ip_route_input() succeeds:
332 * If the output device equals the logical bridge device the packet
333 * came in on, we can consider this bridging. The corresponding MAC
334 * address will be obtained in br_nf_pre_routing_finish_bridge.
335 * Otherwise, the packet is considered to be routed and we just
336 * change the destination MAC address so that the packet will
337 * later be passed up to the IP stack to be routed. For a redirected
338 * packet, ip_route_input() will give back the localhost as output device,
339 * which differs from the bridge device.
341 * Let's now consider the case that ip_route_input() fails:
343 * This can be because the destination address is martian, in which case
344 * the packet will be dropped.
345 * If IP forwarding is disabled, ip_route_input() will fail, while
346 * ip_route_output_key() can return success. The source
347 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
348 * thinks we're handling a locally generated packet and won't care
349 * if IP forwarding is enabled. If the output device equals the logical bridge
350 * device, we proceed as if ip_route_input() succeeded. If it differs from the
351 * logical bridge port or if ip_route_output_key() fails we drop the packet.
353 static int br_nf_pre_routing_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
355 struct net_device *dev = skb->dev;
356 struct iphdr *iph = ip_hdr(skb);
357 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
361 nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
363 if (nf_bridge->pkt_otherhost) {
364 skb->pkt_type = PACKET_OTHERHOST;
365 nf_bridge->pkt_otherhost = false;
367 nf_bridge->in_prerouting = 0;
368 if (br_nf_ipv4_daddr_was_changed(skb, nf_bridge)) {
369 if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
370 struct in_device *in_dev = __in_dev_get_rcu(dev);
372 /* If err equals -EHOSTUNREACH the error is due to a
373 * martian destination or due to the fact that
374 * forwarding is disabled. For most martian packets,
375 * ip_route_output_key() will fail. It won't fail for 2 types of
376 * martian destinations: loopback destinations and destination
377 * 0.0.0.0. In both cases the packet will be dropped because the
378 * destination is the loopback device and not the bridge. */
379 if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
382 rt = ip_route_output(net, iph->daddr, 0,
383 RT_TOS(iph->tos), 0);
385 /* - Bridged-and-DNAT'ed traffic doesn't
386 * require ip_forwarding. */
387 if (rt->dst.dev == dev) {
389 skb_dst_set(skb, &rt->dst);
398 if (skb_dst(skb)->dev == dev) {
400 skb->dev = nf_bridge->physindev;
401 nf_bridge_update_protocol(skb);
402 nf_bridge_push_encap_header(skb);
403 br_nf_hook_thresh(NF_BR_PRE_ROUTING,
404 net, sk, skb, skb->dev,
406 br_nf_pre_routing_finish_bridge);
409 ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr);
410 skb->pkt_type = PACKET_HOST;
413 rt = bridge_parent_rtable(nf_bridge->physindev);
419 skb_dst_set_noref(skb, &rt->dst);
422 skb->dev = nf_bridge->physindev;
423 nf_bridge_update_protocol(skb);
424 nf_bridge_push_encap_header(skb);
425 br_nf_hook_thresh(NF_BR_PRE_ROUTING, net, sk, skb, skb->dev, NULL,
426 br_handle_frame_finish);
430 static struct net_device *brnf_get_logical_dev(struct sk_buff *skb, const struct net_device *dev)
432 struct net_device *vlan, *br;
434 br = bridge_parent(dev);
435 if (brnf_pass_vlan_indev == 0 || !skb_vlan_tag_present(skb))
438 vlan = __vlan_find_dev_deep_rcu(br, skb->vlan_proto,
439 skb_vlan_tag_get(skb) & VLAN_VID_MASK);
441 return vlan ? vlan : br;
444 /* Some common code for IPv4/IPv6 */
445 struct net_device *setup_pre_routing(struct sk_buff *skb)
447 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
449 if (skb->pkt_type == PACKET_OTHERHOST) {
450 skb->pkt_type = PACKET_HOST;
451 nf_bridge->pkt_otherhost = true;
454 nf_bridge->in_prerouting = 1;
455 nf_bridge->physindev = skb->dev;
456 skb->dev = brnf_get_logical_dev(skb, skb->dev);
458 if (skb->protocol == htons(ETH_P_8021Q))
459 nf_bridge->orig_proto = BRNF_PROTO_8021Q;
460 else if (skb->protocol == htons(ETH_P_PPP_SES))
461 nf_bridge->orig_proto = BRNF_PROTO_PPPOE;
463 /* Must drop socket now because of tproxy. */
468 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
469 * Replicate the checks that IPv4 does on packet reception.
470 * Set skb->dev to the bridge device (i.e. parent of the
471 * receiving device) to make netfilter happy, the REDIRECT
472 * target in particular. Save the original destination IP
473 * address to be able to detect DNAT afterwards. */
474 static unsigned int br_nf_pre_routing(void *priv,
476 const struct nf_hook_state *state)
478 struct nf_bridge_info *nf_bridge;
479 struct net_bridge_port *p;
480 struct net_bridge *br;
481 __u32 len = nf_bridge_encap_header_len(skb);
483 if (unlikely(!pskb_may_pull(skb, len)))
486 p = br_port_get_rcu(state->in);
491 if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb)) {
492 if (!brnf_call_ip6tables && !br->nf_call_ip6tables)
495 nf_bridge_pull_encap_header_rcsum(skb);
496 return br_nf_pre_routing_ipv6(priv, skb, state);
499 if (!brnf_call_iptables && !br->nf_call_iptables)
502 if (!IS_IP(skb) && !IS_VLAN_IP(skb) && !IS_PPPOE_IP(skb))
505 nf_bridge_pull_encap_header_rcsum(skb);
507 if (br_validate_ipv4(state->net, skb))
510 nf_bridge_put(skb->nf_bridge);
511 if (!nf_bridge_alloc(skb))
513 if (!setup_pre_routing(skb))
516 nf_bridge = nf_bridge_info_get(skb);
517 nf_bridge->ipv4_daddr = ip_hdr(skb)->daddr;
519 skb->protocol = htons(ETH_P_IP);
520 skb->transport_header = skb->network_header + ip_hdr(skb)->ihl * 4;
522 NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, state->net, state->sk, skb,
524 br_nf_pre_routing_finish);
530 /* PF_BRIDGE/FORWARD *************************************************/
531 static int br_nf_forward_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
533 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
534 struct net_device *in;
536 if (!IS_ARP(skb) && !IS_VLAN_ARP(skb)) {
538 if (skb->protocol == htons(ETH_P_IP))
539 nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
541 if (skb->protocol == htons(ETH_P_IPV6))
542 nf_bridge->frag_max_size = IP6CB(skb)->frag_max_size;
544 in = nf_bridge->physindev;
545 if (nf_bridge->pkt_otherhost) {
546 skb->pkt_type = PACKET_OTHERHOST;
547 nf_bridge->pkt_otherhost = false;
549 nf_bridge_update_protocol(skb);
551 in = *((struct net_device **)(skb->cb));
553 nf_bridge_push_encap_header(skb);
555 br_nf_hook_thresh(NF_BR_FORWARD, net, sk, skb, in, skb->dev,
561 /* This is the 'purely bridged' case. For IP, we pass the packet to
562 * netfilter with indev and outdev set to the bridge device,
563 * but we are still able to filter on the 'real' indev/outdev
564 * because of the physdev module. For ARP, indev and outdev are the
566 static unsigned int br_nf_forward_ip(void *priv,
568 const struct nf_hook_state *state)
570 struct nf_bridge_info *nf_bridge;
571 struct net_device *parent;
577 /* Need exclusive nf_bridge_info since we might have multiple
578 * different physoutdevs. */
579 if (!nf_bridge_unshare(skb))
582 nf_bridge = nf_bridge_info_get(skb);
586 parent = bridge_parent(state->out);
590 if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
592 else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
597 nf_bridge_pull_encap_header(skb);
599 if (skb->pkt_type == PACKET_OTHERHOST) {
600 skb->pkt_type = PACKET_HOST;
601 nf_bridge->pkt_otherhost = true;
604 if (pf == NFPROTO_IPV4) {
605 if (br_validate_ipv4(state->net, skb))
607 IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
610 if (pf == NFPROTO_IPV6) {
611 if (br_validate_ipv6(state->net, skb))
613 IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
616 nf_bridge->physoutdev = skb->dev;
617 if (pf == NFPROTO_IPV4)
618 skb->protocol = htons(ETH_P_IP);
620 skb->protocol = htons(ETH_P_IPV6);
622 NF_HOOK(pf, NF_INET_FORWARD, state->net, NULL, skb,
623 brnf_get_logical_dev(skb, state->in),
624 parent, br_nf_forward_finish);
629 static unsigned int br_nf_forward_arp(void *priv,
631 const struct nf_hook_state *state)
633 struct net_bridge_port *p;
634 struct net_bridge *br;
635 struct net_device **d = (struct net_device **)(skb->cb);
637 p = br_port_get_rcu(state->out);
642 if (!brnf_call_arptables && !br->nf_call_arptables)
646 if (!IS_VLAN_ARP(skb))
648 nf_bridge_pull_encap_header(skb);
651 if (unlikely(!pskb_may_pull(skb, sizeof(struct arphdr))))
654 if (arp_hdr(skb)->ar_pln != 4) {
655 if (IS_VLAN_ARP(skb))
656 nf_bridge_push_encap_header(skb);
660 NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, state->net, state->sk, skb,
661 state->in, state->out, br_nf_forward_finish);
666 static int br_nf_push_frag_xmit(struct net *net, struct sock *sk, struct sk_buff *skb)
668 struct brnf_frag_data *data;
671 data = this_cpu_ptr(&brnf_frag_data_storage);
672 err = skb_cow_head(skb, data->size);
679 if (data->vlan_tci) {
680 skb->vlan_tci = data->vlan_tci;
681 skb->vlan_proto = data->vlan_proto;
684 skb_copy_to_linear_data_offset(skb, -data->size, data->mac, data->size);
685 __skb_push(skb, data->encap_size);
687 nf_bridge_info_free(skb);
688 return br_dev_queue_push_xmit(net, sk, skb);
692 br_nf_ip_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
693 int (*output)(struct net *, struct sock *, struct sk_buff *))
695 unsigned int mtu = ip_skb_dst_mtu(sk, skb);
696 struct iphdr *iph = ip_hdr(skb);
698 if (unlikely(((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) ||
699 (IPCB(skb)->frag_max_size &&
700 IPCB(skb)->frag_max_size > mtu))) {
701 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
706 return ip_do_fragment(net, sk, skb, output);
709 static unsigned int nf_bridge_mtu_reduction(const struct sk_buff *skb)
711 if (skb->nf_bridge->orig_proto == BRNF_PROTO_PPPOE)
712 return PPPOE_SES_HLEN;
716 static int br_nf_dev_queue_xmit(struct net *net, struct sock *sk, struct sk_buff *skb)
718 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
719 unsigned int mtu, mtu_reserved;
721 mtu_reserved = nf_bridge_mtu_reduction(skb);
724 if (nf_bridge->pkt_otherhost) {
725 skb->pkt_type = PACKET_OTHERHOST;
726 nf_bridge->pkt_otherhost = false;
729 if (nf_bridge->frag_max_size && nf_bridge->frag_max_size < mtu)
730 mtu = nf_bridge->frag_max_size;
732 nf_bridge_update_protocol(skb);
733 nf_bridge_push_encap_header(skb);
735 if (skb_is_gso(skb) || skb->len + mtu_reserved <= mtu) {
736 nf_bridge_info_free(skb);
737 return br_dev_queue_push_xmit(net, sk, skb);
740 /* This is wrong! We should preserve the original fragment
741 * boundaries by preserving frag_list rather than refragmenting.
743 if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV4) &&
744 skb->protocol == htons(ETH_P_IP)) {
745 struct brnf_frag_data *data;
747 if (br_validate_ipv4(net, skb))
750 IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
752 data = this_cpu_ptr(&brnf_frag_data_storage);
754 data->vlan_tci = skb->vlan_tci;
755 data->vlan_proto = skb->vlan_proto;
756 data->encap_size = nf_bridge_encap_header_len(skb);
757 data->size = ETH_HLEN + data->encap_size;
759 skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
762 return br_nf_ip_fragment(net, sk, skb, br_nf_push_frag_xmit);
764 if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV6) &&
765 skb->protocol == htons(ETH_P_IPV6)) {
766 const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops();
767 struct brnf_frag_data *data;
769 if (br_validate_ipv6(net, skb))
772 IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
774 data = this_cpu_ptr(&brnf_frag_data_storage);
775 data->encap_size = nf_bridge_encap_header_len(skb);
776 data->size = ETH_HLEN + data->encap_size;
778 skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
782 return v6ops->fragment(net, sk, skb, br_nf_push_frag_xmit);
787 nf_bridge_info_free(skb);
788 return br_dev_queue_push_xmit(net, sk, skb);
794 /* PF_BRIDGE/POST_ROUTING ********************************************/
795 static unsigned int br_nf_post_routing(void *priv,
797 const struct nf_hook_state *state)
799 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
800 struct net_device *realoutdev = bridge_parent(skb->dev);
803 /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
804 * on a bridge, but was delivered locally and is now being routed:
806 * POST_ROUTING was already invoked from the ip stack.
808 if (!nf_bridge || !nf_bridge->physoutdev)
814 if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
816 else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
821 if (skb->pkt_type == PACKET_OTHERHOST) {
822 skb->pkt_type = PACKET_HOST;
823 nf_bridge->pkt_otherhost = true;
826 nf_bridge_pull_encap_header(skb);
827 if (pf == NFPROTO_IPV4)
828 skb->protocol = htons(ETH_P_IP);
830 skb->protocol = htons(ETH_P_IPV6);
832 NF_HOOK(pf, NF_INET_POST_ROUTING, state->net, state->sk, skb,
834 br_nf_dev_queue_xmit);
839 /* IP/SABOTAGE *****************************************************/
840 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
841 * for the second time. */
842 static unsigned int ip_sabotage_in(void *priv,
844 const struct nf_hook_state *state)
846 if (skb->nf_bridge && !skb->nf_bridge->in_prerouting &&
847 !netif_is_l3_master(skb->dev)) {
848 state->okfn(state->net, state->sk, skb);
855 /* This is called when br_netfilter has called into iptables/netfilter,
856 * and DNAT has taken place on a bridge-forwarded packet.
858 * neigh->output has created a new MAC header, with local br0 MAC
861 * This restores the original MAC saddr of the bridged packet
862 * before invoking bridge forward logic to transmit the packet.
864 static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff *skb)
866 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
868 skb_pull(skb, ETH_HLEN);
869 nf_bridge->bridged_dnat = 0;
871 BUILD_BUG_ON(sizeof(nf_bridge->neigh_header) != (ETH_HLEN - ETH_ALEN));
873 skb_copy_to_linear_data_offset(skb, -(ETH_HLEN - ETH_ALEN),
874 nf_bridge->neigh_header,
875 ETH_HLEN - ETH_ALEN);
876 skb->dev = nf_bridge->physindev;
878 nf_bridge->physoutdev = NULL;
879 br_handle_frame_finish(dev_net(skb->dev), NULL, skb);
882 static int br_nf_dev_xmit(struct sk_buff *skb)
884 if (skb->nf_bridge && skb->nf_bridge->bridged_dnat) {
885 br_nf_pre_routing_finish_bridge_slow(skb);
891 static const struct nf_br_ops br_ops = {
892 .br_dev_xmit_hook = br_nf_dev_xmit,
895 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
896 * br_dev_queue_push_xmit is called afterwards */
897 static const struct nf_hook_ops br_nf_ops[] = {
899 .hook = br_nf_pre_routing,
900 .pf = NFPROTO_BRIDGE,
901 .hooknum = NF_BR_PRE_ROUTING,
902 .priority = NF_BR_PRI_BRNF,
905 .hook = br_nf_forward_ip,
906 .pf = NFPROTO_BRIDGE,
907 .hooknum = NF_BR_FORWARD,
908 .priority = NF_BR_PRI_BRNF - 1,
911 .hook = br_nf_forward_arp,
912 .pf = NFPROTO_BRIDGE,
913 .hooknum = NF_BR_FORWARD,
914 .priority = NF_BR_PRI_BRNF,
917 .hook = br_nf_post_routing,
918 .pf = NFPROTO_BRIDGE,
919 .hooknum = NF_BR_POST_ROUTING,
920 .priority = NF_BR_PRI_LAST,
923 .hook = ip_sabotage_in,
925 .hooknum = NF_INET_PRE_ROUTING,
926 .priority = NF_IP_PRI_FIRST,
929 .hook = ip_sabotage_in,
931 .hooknum = NF_INET_PRE_ROUTING,
932 .priority = NF_IP6_PRI_FIRST,
936 static int brnf_device_event(struct notifier_block *unused, unsigned long event,
939 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
940 struct brnf_net *brnet;
944 if (event != NETDEV_REGISTER || !(dev->priv_flags & IFF_EBRIDGE))
950 brnet = net_generic(net, brnf_net_id);
954 ret = nf_register_net_hooks(net, br_nf_ops, ARRAY_SIZE(br_nf_ops));
958 brnet->enabled = true;
962 static void __net_exit brnf_exit_net(struct net *net)
964 struct brnf_net *brnet = net_generic(net, brnf_net_id);
969 nf_unregister_net_hooks(net, br_nf_ops, ARRAY_SIZE(br_nf_ops));
970 brnet->enabled = false;
973 static struct pernet_operations brnf_net_ops __read_mostly = {
974 .exit = brnf_exit_net,
976 .size = sizeof(struct brnf_net),
979 static struct notifier_block brnf_notifier __read_mostly = {
980 .notifier_call = brnf_device_event,
983 /* recursively invokes nf_hook_slow (again), skipping already-called
984 * hooks (< NF_BR_PRI_BRNF).
986 * Called with rcu read lock held.
988 int br_nf_hook_thresh(unsigned int hook, struct net *net,
989 struct sock *sk, struct sk_buff *skb,
990 struct net_device *indev,
991 struct net_device *outdev,
992 int (*okfn)(struct net *, struct sock *,
995 const struct nf_hook_entries *e;
996 struct nf_hook_state state;
997 struct nf_hook_ops **ops;
1001 e = rcu_dereference(net->nf.hooks_bridge[hook]);
1003 return okfn(net, sk, skb);
1005 ops = nf_hook_entries_get_hook_ops(e);
1006 for (i = 0; i < e->num_hook_entries; i++) {
1007 /* These hooks have already been called */
1008 if (ops[i]->priority < NF_BR_PRI_BRNF)
1011 /* These hooks have not been called yet, run them. */
1012 if (ops[i]->priority > NF_BR_PRI_BRNF)
1015 /* take a closer look at NF_BR_PRI_BRNF. */
1016 if (ops[i]->hook == br_nf_pre_routing) {
1017 /* This hook diverted the skb to this function,
1018 * hooks after this have not been run yet.
1025 nf_hook_state_init(&state, hook, NFPROTO_BRIDGE, indev, outdev,
1028 ret = nf_hook_slow(skb, &state, e, i);
1030 ret = okfn(net, sk, skb);
1035 #ifdef CONFIG_SYSCTL
1037 int brnf_sysctl_call_tables(struct ctl_table *ctl, int write,
1038 void __user *buffer, size_t *lenp, loff_t *ppos)
1042 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1044 if (write && *(int *)(ctl->data))
1045 *(int *)(ctl->data) = 1;
1049 static struct ctl_table brnf_table[] = {
1051 .procname = "bridge-nf-call-arptables",
1052 .data = &brnf_call_arptables,
1053 .maxlen = sizeof(int),
1055 .proc_handler = brnf_sysctl_call_tables,
1058 .procname = "bridge-nf-call-iptables",
1059 .data = &brnf_call_iptables,
1060 .maxlen = sizeof(int),
1062 .proc_handler = brnf_sysctl_call_tables,
1065 .procname = "bridge-nf-call-ip6tables",
1066 .data = &brnf_call_ip6tables,
1067 .maxlen = sizeof(int),
1069 .proc_handler = brnf_sysctl_call_tables,
1072 .procname = "bridge-nf-filter-vlan-tagged",
1073 .data = &brnf_filter_vlan_tagged,
1074 .maxlen = sizeof(int),
1076 .proc_handler = brnf_sysctl_call_tables,
1079 .procname = "bridge-nf-filter-pppoe-tagged",
1080 .data = &brnf_filter_pppoe_tagged,
1081 .maxlen = sizeof(int),
1083 .proc_handler = brnf_sysctl_call_tables,
1086 .procname = "bridge-nf-pass-vlan-input-dev",
1087 .data = &brnf_pass_vlan_indev,
1088 .maxlen = sizeof(int),
1090 .proc_handler = brnf_sysctl_call_tables,
1096 static int __init br_netfilter_init(void)
1100 ret = register_pernet_subsys(&brnf_net_ops);
1104 ret = register_netdevice_notifier(&brnf_notifier);
1106 unregister_pernet_subsys(&brnf_net_ops);
1110 #ifdef CONFIG_SYSCTL
1111 brnf_sysctl_header = register_net_sysctl(&init_net, "net/bridge", brnf_table);
1112 if (brnf_sysctl_header == NULL) {
1114 "br_netfilter: can't register to sysctl.\n");
1115 unregister_netdevice_notifier(&brnf_notifier);
1116 unregister_pernet_subsys(&brnf_net_ops);
1120 RCU_INIT_POINTER(nf_br_ops, &br_ops);
1121 printk(KERN_NOTICE "Bridge firewalling registered\n");
1125 static void __exit br_netfilter_fini(void)
1127 RCU_INIT_POINTER(nf_br_ops, NULL);
1128 unregister_netdevice_notifier(&brnf_notifier);
1129 unregister_pernet_subsys(&brnf_net_ops);
1130 #ifdef CONFIG_SYSCTL
1131 unregister_net_sysctl_table(brnf_sysctl_header);
1135 module_init(br_netfilter_init);
1136 module_exit(br_netfilter_fini);
1138 MODULE_LICENSE("GPL");
1139 MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
1140 MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
1141 MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");