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>
20 #include <linux/netdevice.h>
21 #include <linux/skbuff.h>
22 #include <linux/if_arp.h>
23 #include <linux/if_ether.h>
24 #include <linux/if_vlan.h>
25 #include <linux/if_pppox.h>
26 #include <linux/ppp_defs.h>
27 #include <linux/netfilter_bridge.h>
28 #include <linux/netfilter_ipv4.h>
29 #include <linux/netfilter_ipv6.h>
30 #include <linux/netfilter_arp.h>
31 #include <linux/in_route.h>
32 #include <linux/inetdevice.h>
36 #include <net/route.h>
38 #include <asm/uaccess.h>
39 #include "br_private.h"
41 #include <linux/sysctl.h>
44 #define skb_origaddr(skb) (((struct bridge_skb_cb *) \
45 (skb->nf_bridge->data))->daddr.ipv4)
46 #define store_orig_dstaddr(skb) (skb_origaddr(skb) = ip_hdr(skb)->daddr)
47 #define dnat_took_place(skb) (skb_origaddr(skb) != ip_hdr(skb)->daddr)
50 static struct ctl_table_header *brnf_sysctl_header;
51 static int brnf_call_iptables __read_mostly = 1;
52 static int brnf_call_ip6tables __read_mostly = 1;
53 static int brnf_call_arptables __read_mostly = 1;
54 static int brnf_filter_vlan_tagged __read_mostly = 0;
55 static int brnf_filter_pppoe_tagged __read_mostly = 0;
57 #define brnf_filter_vlan_tagged 0
58 #define brnf_filter_pppoe_tagged 0
61 static inline __be16 vlan_proto(const struct sk_buff *skb)
63 return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
66 #define IS_VLAN_IP(skb) \
67 (skb->protocol == htons(ETH_P_8021Q) && \
68 vlan_proto(skb) == htons(ETH_P_IP) && \
69 brnf_filter_vlan_tagged)
71 #define IS_VLAN_IPV6(skb) \
72 (skb->protocol == htons(ETH_P_8021Q) && \
73 vlan_proto(skb) == htons(ETH_P_IPV6) &&\
74 brnf_filter_vlan_tagged)
76 #define IS_VLAN_ARP(skb) \
77 (skb->protocol == htons(ETH_P_8021Q) && \
78 vlan_proto(skb) == htons(ETH_P_ARP) && \
79 brnf_filter_vlan_tagged)
81 static inline __be16 pppoe_proto(const struct sk_buff *skb)
83 return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
84 sizeof(struct pppoe_hdr)));
87 #define IS_PPPOE_IP(skb) \
88 (skb->protocol == htons(ETH_P_PPP_SES) && \
89 pppoe_proto(skb) == htons(PPP_IP) && \
90 brnf_filter_pppoe_tagged)
92 #define IS_PPPOE_IPV6(skb) \
93 (skb->protocol == htons(ETH_P_PPP_SES) && \
94 pppoe_proto(skb) == htons(PPP_IPV6) && \
95 brnf_filter_pppoe_tagged)
97 static void fake_update_pmtu(struct dst_entry *dst, u32 mtu)
101 static struct dst_ops fake_dst_ops = {
103 .protocol = cpu_to_be16(ETH_P_IP),
104 .update_pmtu = fake_update_pmtu,
105 .entries = ATOMIC_INIT(0),
109 * Initialize bogus route table used to keep netfilter happy.
110 * Currently, we fill in the PMTU entry because netfilter
111 * refragmentation needs it, and the rt_flags entry because
112 * ipt_REJECT needs it. Future netfilter modules might
113 * require us to fill additional fields.
115 void br_netfilter_rtable_init(struct net_bridge *br)
117 struct rtable *rt = &br->fake_rtable;
119 atomic_set(&rt->u.dst.__refcnt, 1);
120 rt->u.dst.dev = br->dev;
121 rt->u.dst.path = &rt->u.dst;
122 rt->u.dst.metrics[RTAX_MTU - 1] = 1500;
123 rt->u.dst.flags = DST_NOXFRM;
124 rt->u.dst.ops = &fake_dst_ops;
127 static inline struct rtable *bridge_parent_rtable(const struct net_device *dev)
129 struct net_bridge_port *port = rcu_dereference(dev->br_port);
131 return port ? &port->br->fake_rtable : NULL;
134 static inline struct net_device *bridge_parent(const struct net_device *dev)
136 struct net_bridge_port *port = rcu_dereference(dev->br_port);
138 return port ? port->br->dev : NULL;
141 static inline struct nf_bridge_info *nf_bridge_alloc(struct sk_buff *skb)
143 skb->nf_bridge = kzalloc(sizeof(struct nf_bridge_info), GFP_ATOMIC);
144 if (likely(skb->nf_bridge))
145 atomic_set(&(skb->nf_bridge->use), 1);
147 return skb->nf_bridge;
150 static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
152 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
154 if (atomic_read(&nf_bridge->use) > 1) {
155 struct nf_bridge_info *tmp = nf_bridge_alloc(skb);
158 memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info));
159 atomic_set(&tmp->use, 1);
160 nf_bridge_put(nf_bridge);
167 static inline void nf_bridge_push_encap_header(struct sk_buff *skb)
169 unsigned int len = nf_bridge_encap_header_len(skb);
172 skb->network_header -= len;
175 static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
177 unsigned int len = nf_bridge_encap_header_len(skb);
180 skb->network_header += len;
183 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
185 unsigned int len = nf_bridge_encap_header_len(skb);
187 skb_pull_rcsum(skb, len);
188 skb->network_header += len;
191 static inline void nf_bridge_save_header(struct sk_buff *skb)
193 int header_size = ETH_HLEN + nf_bridge_encap_header_len(skb);
195 skb_copy_from_linear_data_offset(skb, -header_size,
196 skb->nf_bridge->data, header_size);
200 * When forwarding bridge frames, we save a copy of the original
201 * header before processing.
203 int nf_bridge_copy_header(struct sk_buff *skb)
206 int header_size = ETH_HLEN + nf_bridge_encap_header_len(skb);
208 err = skb_cow_head(skb, header_size);
212 skb_copy_to_linear_data_offset(skb, -header_size,
213 skb->nf_bridge->data, header_size);
214 __skb_push(skb, nf_bridge_encap_header_len(skb));
218 /* PF_BRIDGE/PRE_ROUTING *********************************************/
219 /* Undo the changes made for ip6tables PREROUTING and continue the
220 * bridge PRE_ROUTING hook. */
221 static int br_nf_pre_routing_finish_ipv6(struct sk_buff *skb)
223 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
226 if (nf_bridge->mask & BRNF_PKT_TYPE) {
227 skb->pkt_type = PACKET_OTHERHOST;
228 nf_bridge->mask ^= BRNF_PKT_TYPE;
230 nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
232 rt = bridge_parent_rtable(nf_bridge->physindev);
237 dst_hold(&rt->u.dst);
238 skb_dst_set(skb, &rt->u.dst);
240 skb->dev = nf_bridge->physindev;
241 nf_bridge_push_encap_header(skb);
242 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
243 br_handle_frame_finish, 1);
248 /* This requires some explaining. If DNAT has taken place,
249 * we will need to fix up the destination Ethernet address,
250 * and this is a tricky process.
252 * There are two cases to consider:
253 * 1. The packet was DNAT'ed to a device in the same bridge
254 * port group as it was received on. We can still bridge
256 * 2. The packet was DNAT'ed to a different device, either
257 * a non-bridged device or another bridge port group.
258 * The packet will need to be routed.
260 * The correct way of distinguishing between these two cases is to
261 * call ip_route_input() and to look at skb->dst->dev, which is
262 * changed to the destination device if ip_route_input() succeeds.
264 * Let us first consider the case that ip_route_input() succeeds:
266 * If skb->dst->dev equals the logical bridge device the packet
267 * came in on, we can consider this bridging. The packet is passed
268 * through the neighbour output function to build a new destination
269 * MAC address, which will make the packet enter br_nf_local_out()
270 * not much later. In that function it is assured that the iptables
271 * FORWARD chain is traversed for the packet.
273 * Otherwise, the packet is considered to be routed and we just
274 * change the destination MAC address so that the packet will
275 * later be passed up to the IP stack to be routed. For a redirected
276 * packet, ip_route_input() will give back the localhost as output device,
277 * which differs from the bridge device.
279 * Let us now consider the case that ip_route_input() fails:
281 * This can be because the destination address is martian, in which case
282 * the packet will be dropped.
283 * After a "echo '0' > /proc/sys/net/ipv4/ip_forward" ip_route_input()
284 * will fail, while __ip_route_output_key() will return success. The source
285 * address for __ip_route_output_key() is set to zero, so __ip_route_output_key
286 * thinks we're handling a locally generated packet and won't care
287 * if IP forwarding is allowed. We send a warning message to the users's
288 * log telling her to put IP forwarding on.
290 * ip_route_input() will also fail if there is no route available.
291 * In that case we just drop the packet.
293 * --Lennert, 20020411
294 * --Bart, 20020416 (updated)
295 * --Bart, 20021007 (updated)
296 * --Bart, 20062711 (updated) */
297 static int br_nf_pre_routing_finish_bridge(struct sk_buff *skb)
299 if (skb->pkt_type == PACKET_OTHERHOST) {
300 skb->pkt_type = PACKET_HOST;
301 skb->nf_bridge->mask |= BRNF_PKT_TYPE;
303 skb->nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
305 skb->dev = bridge_parent(skb->dev);
307 struct dst_entry *dst = skb_dst(skb);
309 nf_bridge_pull_encap_header(skb);
312 return neigh_hh_output(dst->hh, skb);
313 else if (dst->neighbour)
314 return dst->neighbour->output(skb);
320 static int br_nf_pre_routing_finish(struct sk_buff *skb)
322 struct net_device *dev = skb->dev;
323 struct iphdr *iph = ip_hdr(skb);
324 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
328 if (nf_bridge->mask & BRNF_PKT_TYPE) {
329 skb->pkt_type = PACKET_OTHERHOST;
330 nf_bridge->mask ^= BRNF_PKT_TYPE;
332 nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
333 if (dnat_took_place(skb)) {
334 if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
340 .tos = RT_TOS(iph->tos) },
344 struct in_device *in_dev = __in_dev_get_rcu(dev);
346 /* If err equals -EHOSTUNREACH the error is due to a
347 * martian destination or due to the fact that
348 * forwarding is disabled. For most martian packets,
349 * ip_route_output_key() will fail. It won't fail for 2 types of
350 * martian destinations: loopback destinations and destination
351 * 0.0.0.0. In both cases the packet will be dropped because the
352 * destination is the loopback device and not the bridge. */
353 if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
356 if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
357 /* - Bridged-and-DNAT'ed traffic doesn't
358 * require ip_forwarding. */
359 if (((struct dst_entry *)rt)->dev == dev) {
360 skb_dst_set(skb, (struct dst_entry *)rt);
363 dst_release((struct dst_entry *)rt);
369 if (skb_dst(skb)->dev == dev) {
371 /* Tell br_nf_local_out this is a
373 nf_bridge->mask |= BRNF_BRIDGED_DNAT;
374 skb->dev = nf_bridge->physindev;
375 nf_bridge_push_encap_header(skb);
376 NF_HOOK_THRESH(NFPROTO_BRIDGE,
379 br_nf_pre_routing_finish_bridge,
383 memcpy(eth_hdr(skb)->h_dest, dev->dev_addr, ETH_ALEN);
384 skb->pkt_type = PACKET_HOST;
387 rt = bridge_parent_rtable(nf_bridge->physindev);
392 dst_hold(&rt->u.dst);
393 skb_dst_set(skb, &rt->u.dst);
396 skb->dev = nf_bridge->physindev;
397 nf_bridge_push_encap_header(skb);
398 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
399 br_handle_frame_finish, 1);
404 /* Some common code for IPv4/IPv6 */
405 static struct net_device *setup_pre_routing(struct sk_buff *skb)
407 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
409 if (skb->pkt_type == PACKET_OTHERHOST) {
410 skb->pkt_type = PACKET_HOST;
411 nf_bridge->mask |= BRNF_PKT_TYPE;
414 nf_bridge->mask |= BRNF_NF_BRIDGE_PREROUTING;
415 nf_bridge->physindev = skb->dev;
416 skb->dev = bridge_parent(skb->dev);
421 /* We only check the length. A bridge shouldn't do any hop-by-hop stuff anyway */
422 static int check_hbh_len(struct sk_buff *skb)
424 unsigned char *raw = (u8 *)(ipv6_hdr(skb) + 1);
426 const unsigned char *nh = skb_network_header(skb);
428 int len = (raw[1] + 1) << 3;
430 if ((raw + len) - skb->data > skb_headlen(skb))
437 int optlen = nh[off + 1] + 2;
448 if (nh[off + 1] != 4 || (off & 3) != 2)
450 pkt_len = ntohl(*(__be32 *) (nh + off + 2));
451 if (pkt_len <= IPV6_MAXPLEN ||
452 ipv6_hdr(skb)->payload_len)
454 if (pkt_len > skb->len - sizeof(struct ipv6hdr))
456 if (pskb_trim_rcsum(skb,
457 pkt_len + sizeof(struct ipv6hdr)))
459 nh = skb_network_header(skb);
476 /* Replicate the checks that IPv6 does on packet reception and pass the packet
477 * to ip6tables, which doesn't support NAT, so things are fairly simple. */
478 static unsigned int br_nf_pre_routing_ipv6(unsigned int hook,
480 const struct net_device *in,
481 const struct net_device *out,
482 int (*okfn)(struct sk_buff *))
487 if (skb->len < sizeof(struct ipv6hdr))
490 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
495 if (hdr->version != 6)
498 pkt_len = ntohs(hdr->payload_len);
500 if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
501 if (pkt_len + sizeof(struct ipv6hdr) > skb->len)
503 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
506 if (hdr->nexthdr == NEXTHDR_HOP && check_hbh_len(skb))
509 nf_bridge_put(skb->nf_bridge);
510 if (!nf_bridge_alloc(skb))
512 if (!setup_pre_routing(skb))
515 NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
516 br_nf_pre_routing_finish_ipv6);
524 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
525 * Replicate the checks that IPv4 does on packet reception.
526 * Set skb->dev to the bridge device (i.e. parent of the
527 * receiving device) to make netfilter happy, the REDIRECT
528 * target in particular. Save the original destination IP
529 * address to be able to detect DNAT afterwards. */
530 static unsigned int br_nf_pre_routing(unsigned int hook, struct sk_buff *skb,
531 const struct net_device *in,
532 const struct net_device *out,
533 int (*okfn)(struct sk_buff *))
536 __u32 len = nf_bridge_encap_header_len(skb);
538 if (unlikely(!pskb_may_pull(skb, len)))
541 if (skb->protocol == htons(ETH_P_IPV6) || IS_VLAN_IPV6(skb) ||
542 IS_PPPOE_IPV6(skb)) {
544 if (!brnf_call_ip6tables)
547 nf_bridge_pull_encap_header_rcsum(skb);
548 return br_nf_pre_routing_ipv6(hook, skb, in, out, okfn);
551 if (!brnf_call_iptables)
555 if (skb->protocol != htons(ETH_P_IP) && !IS_VLAN_IP(skb) &&
559 nf_bridge_pull_encap_header_rcsum(skb);
561 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
565 if (iph->ihl < 5 || iph->version != 4)
568 if (!pskb_may_pull(skb, 4 * iph->ihl))
572 if (ip_fast_csum((__u8 *) iph, iph->ihl) != 0)
575 len = ntohs(iph->tot_len);
576 if (skb->len < len || len < 4 * iph->ihl)
579 pskb_trim_rcsum(skb, len);
581 nf_bridge_put(skb->nf_bridge);
582 if (!nf_bridge_alloc(skb))
584 if (!setup_pre_routing(skb))
586 store_orig_dstaddr(skb);
588 NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
589 br_nf_pre_routing_finish);
594 // IP_INC_STATS_BH(IpInHdrErrors);
600 /* PF_BRIDGE/LOCAL_IN ************************************************/
601 /* The packet is locally destined, which requires a real
602 * dst_entry, so detach the fake one. On the way up, the
603 * packet would pass through PRE_ROUTING again (which already
604 * took place when the packet entered the bridge), but we
605 * register an IPv4 PRE_ROUTING 'sabotage' hook that will
606 * prevent this from happening. */
607 static unsigned int br_nf_local_in(unsigned int hook, struct sk_buff *skb,
608 const struct net_device *in,
609 const struct net_device *out,
610 int (*okfn)(struct sk_buff *))
612 struct rtable *rt = skb_rtable(skb);
614 if (rt && rt == bridge_parent_rtable(in))
620 /* PF_BRIDGE/FORWARD *************************************************/
621 static int br_nf_forward_finish(struct sk_buff *skb)
623 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
624 struct net_device *in;
626 if (skb->protocol != htons(ETH_P_ARP) && !IS_VLAN_ARP(skb)) {
627 in = nf_bridge->physindev;
628 if (nf_bridge->mask & BRNF_PKT_TYPE) {
629 skb->pkt_type = PACKET_OTHERHOST;
630 nf_bridge->mask ^= BRNF_PKT_TYPE;
633 in = *((struct net_device **)(skb->cb));
635 nf_bridge_push_encap_header(skb);
636 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_FORWARD, skb, in,
637 skb->dev, br_forward_finish, 1);
641 /* This is the 'purely bridged' case. For IP, we pass the packet to
642 * netfilter with indev and outdev set to the bridge device,
643 * but we are still able to filter on the 'real' indev/outdev
644 * because of the physdev module. For ARP, indev and outdev are the
646 static unsigned int br_nf_forward_ip(unsigned int hook, struct sk_buff *skb,
647 const struct net_device *in,
648 const struct net_device *out,
649 int (*okfn)(struct sk_buff *))
651 struct nf_bridge_info *nf_bridge;
652 struct net_device *parent;
658 /* Need exclusive nf_bridge_info since we might have multiple
659 * different physoutdevs. */
660 if (!nf_bridge_unshare(skb))
663 parent = bridge_parent(out);
667 if (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb) ||
670 else if (skb->protocol == htons(ETH_P_IPV6) || IS_VLAN_IPV6(skb) ||
676 nf_bridge_pull_encap_header(skb);
678 nf_bridge = skb->nf_bridge;
679 if (skb->pkt_type == PACKET_OTHERHOST) {
680 skb->pkt_type = PACKET_HOST;
681 nf_bridge->mask |= BRNF_PKT_TYPE;
684 /* The physdev module checks on this */
685 nf_bridge->mask |= BRNF_BRIDGED;
686 nf_bridge->physoutdev = skb->dev;
688 NF_HOOK(pf, NF_INET_FORWARD, skb, bridge_parent(in), parent,
689 br_nf_forward_finish);
694 static unsigned int br_nf_forward_arp(unsigned int hook, struct sk_buff *skb,
695 const struct net_device *in,
696 const struct net_device *out,
697 int (*okfn)(struct sk_buff *))
699 struct net_device **d = (struct net_device **)(skb->cb);
702 if (!brnf_call_arptables)
706 if (skb->protocol != htons(ETH_P_ARP)) {
707 if (!IS_VLAN_ARP(skb))
709 nf_bridge_pull_encap_header(skb);
712 if (arp_hdr(skb)->ar_pln != 4) {
713 if (IS_VLAN_ARP(skb))
714 nf_bridge_push_encap_header(skb);
717 *d = (struct net_device *)in;
718 NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, skb, (struct net_device *)in,
719 (struct net_device *)out, br_nf_forward_finish);
724 /* PF_BRIDGE/LOCAL_OUT ***********************************************
726 * This function sees both locally originated IP packets and forwarded
727 * IP packets (in both cases the destination device is a bridge
728 * device). It also sees bridged-and-DNAT'ed packets.
730 * If (nf_bridge->mask & BRNF_BRIDGED_DNAT) then the packet is bridged
731 * and we fake the PF_BRIDGE/FORWARD hook. The function br_nf_forward()
732 * will then fake the PF_INET/FORWARD hook. br_nf_local_out() has priority
733 * NF_BR_PRI_FIRST, so no relevant PF_BRIDGE/INPUT functions have been nor
736 static unsigned int br_nf_local_out(unsigned int hook, struct sk_buff *skb,
737 const struct net_device *in,
738 const struct net_device *out,
739 int (*okfn)(struct sk_buff *))
741 struct net_device *realindev;
742 struct nf_bridge_info *nf_bridge;
747 /* Need exclusive nf_bridge_info since we might have multiple
748 * different physoutdevs. */
749 if (!nf_bridge_unshare(skb))
752 nf_bridge = skb->nf_bridge;
753 if (!(nf_bridge->mask & BRNF_BRIDGED_DNAT))
756 /* Bridged, take PF_BRIDGE/FORWARD.
757 * (see big note in front of br_nf_pre_routing_finish) */
758 nf_bridge->physoutdev = skb->dev;
759 realindev = nf_bridge->physindev;
761 if (nf_bridge->mask & BRNF_PKT_TYPE) {
762 skb->pkt_type = PACKET_OTHERHOST;
763 nf_bridge->mask ^= BRNF_PKT_TYPE;
765 nf_bridge_push_encap_header(skb);
767 NF_HOOK(NFPROTO_BRIDGE, NF_BR_FORWARD, skb, realindev, skb->dev,
772 #if defined(CONFIG_NF_CONNTRACK_IPV4) || defined(CONFIG_NF_CONNTRACK_IPV4_MODULE)
773 static int br_nf_dev_queue_xmit(struct sk_buff *skb)
775 if (skb->nfct != NULL &&
776 (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb)) &&
777 skb->len > skb->dev->mtu &&
779 return ip_fragment(skb, br_dev_queue_push_xmit);
781 return br_dev_queue_push_xmit(skb);
784 static int br_nf_dev_queue_xmit(struct sk_buff *skb)
786 return br_dev_queue_push_xmit(skb);
790 /* PF_BRIDGE/POST_ROUTING ********************************************/
791 static unsigned int br_nf_post_routing(unsigned int hook, struct sk_buff *skb,
792 const struct net_device *in,
793 const struct net_device *out,
794 int (*okfn)(struct sk_buff *))
796 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
797 struct net_device *realoutdev = bridge_parent(skb->dev);
803 if (!(nf_bridge->mask & (BRNF_BRIDGED | BRNF_BRIDGED_DNAT)))
809 if (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb) ||
812 else if (skb->protocol == htons(ETH_P_IPV6) || IS_VLAN_IPV6(skb) ||
818 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
819 * about the value of skb->pkt_type. */
820 if (skb->pkt_type == PACKET_OTHERHOST) {
821 skb->pkt_type = PACKET_HOST;
822 nf_bridge->mask |= BRNF_PKT_TYPE;
825 nf_bridge_pull_encap_header(skb);
826 nf_bridge_save_header(skb);
828 NF_HOOK(pf, NF_INET_POST_ROUTING, skb, NULL, realoutdev,
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(unsigned int hook, struct sk_buff *skb,
838 const struct net_device *in,
839 const struct net_device *out,
840 int (*okfn)(struct sk_buff *))
842 if (skb->nf_bridge &&
843 !(skb->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)) {
850 /* For br_nf_local_out we need (prio = NF_BR_PRI_FIRST), to insure that innocent
851 * PF_BRIDGE/NF_BR_LOCAL_OUT functions don't get bridged traffic as input.
852 * For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
853 * ip_refrag() can return NF_STOLEN. */
854 static struct nf_hook_ops br_nf_ops[] __read_mostly = {
856 .hook = br_nf_pre_routing,
857 .owner = THIS_MODULE,
859 .hooknum = NF_BR_PRE_ROUTING,
860 .priority = NF_BR_PRI_BRNF,
863 .hook = br_nf_local_in,
864 .owner = THIS_MODULE,
866 .hooknum = NF_BR_LOCAL_IN,
867 .priority = NF_BR_PRI_BRNF,
870 .hook = br_nf_forward_ip,
871 .owner = THIS_MODULE,
873 .hooknum = NF_BR_FORWARD,
874 .priority = NF_BR_PRI_BRNF - 1,
877 .hook = br_nf_forward_arp,
878 .owner = THIS_MODULE,
880 .hooknum = NF_BR_FORWARD,
881 .priority = NF_BR_PRI_BRNF,
884 .hook = br_nf_local_out,
885 .owner = THIS_MODULE,
887 .hooknum = NF_BR_LOCAL_OUT,
888 .priority = NF_BR_PRI_FIRST,
891 .hook = br_nf_post_routing,
892 .owner = THIS_MODULE,
894 .hooknum = NF_BR_POST_ROUTING,
895 .priority = NF_BR_PRI_LAST,
898 .hook = ip_sabotage_in,
899 .owner = THIS_MODULE,
901 .hooknum = NF_INET_PRE_ROUTING,
902 .priority = NF_IP_PRI_FIRST,
905 .hook = ip_sabotage_in,
906 .owner = THIS_MODULE,
908 .hooknum = NF_INET_PRE_ROUTING,
909 .priority = NF_IP6_PRI_FIRST,
915 int brnf_sysctl_call_tables(ctl_table * ctl, int write,
916 void __user * buffer, size_t * lenp, loff_t * ppos)
920 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
922 if (write && *(int *)(ctl->data))
923 *(int *)(ctl->data) = 1;
927 static ctl_table brnf_table[] = {
929 .procname = "bridge-nf-call-arptables",
930 .data = &brnf_call_arptables,
931 .maxlen = sizeof(int),
933 .proc_handler = brnf_sysctl_call_tables,
936 .procname = "bridge-nf-call-iptables",
937 .data = &brnf_call_iptables,
938 .maxlen = sizeof(int),
940 .proc_handler = brnf_sysctl_call_tables,
943 .procname = "bridge-nf-call-ip6tables",
944 .data = &brnf_call_ip6tables,
945 .maxlen = sizeof(int),
947 .proc_handler = brnf_sysctl_call_tables,
950 .procname = "bridge-nf-filter-vlan-tagged",
951 .data = &brnf_filter_vlan_tagged,
952 .maxlen = sizeof(int),
954 .proc_handler = brnf_sysctl_call_tables,
957 .procname = "bridge-nf-filter-pppoe-tagged",
958 .data = &brnf_filter_pppoe_tagged,
959 .maxlen = sizeof(int),
961 .proc_handler = brnf_sysctl_call_tables,
966 static struct ctl_path brnf_path[] = {
967 { .procname = "net", },
968 { .procname = "bridge", },
973 int __init br_netfilter_init(void)
977 ret = nf_register_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
981 brnf_sysctl_header = register_sysctl_paths(brnf_path, brnf_table);
982 if (brnf_sysctl_header == NULL) {
984 "br_netfilter: can't register to sysctl.\n");
985 nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
989 printk(KERN_NOTICE "Bridge firewalling registered\n");
993 void br_netfilter_fini(void)
995 nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
997 unregister_sysctl_table(brnf_sysctl_header);