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/inetdevice.h>
37 #include <net/route.h>
39 #include <asm/uaccess.h>
40 #include "br_private.h"
42 #include <linux/sysctl.h>
45 #define skb_origaddr(skb) (((struct bridge_skb_cb *) \
46 (skb->nf_bridge->data))->daddr.ipv4)
47 #define store_orig_dstaddr(skb) (skb_origaddr(skb) = ip_hdr(skb)->daddr)
48 #define dnat_took_place(skb) (skb_origaddr(skb) != ip_hdr(skb)->daddr)
51 static struct ctl_table_header *brnf_sysctl_header;
52 static int brnf_call_iptables __read_mostly = 1;
53 static int brnf_call_ip6tables __read_mostly = 1;
54 static int brnf_call_arptables __read_mostly = 1;
55 static int brnf_filter_vlan_tagged __read_mostly = 0;
56 static int brnf_filter_pppoe_tagged __read_mostly = 0;
58 #define brnf_call_iptables 1
59 #define brnf_call_ip6tables 1
60 #define brnf_call_arptables 1
61 #define brnf_filter_vlan_tagged 0
62 #define brnf_filter_pppoe_tagged 0
65 static inline __be16 vlan_proto(const struct sk_buff *skb)
67 return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
70 #define IS_VLAN_IP(skb) \
71 (skb->protocol == htons(ETH_P_8021Q) && \
72 vlan_proto(skb) == htons(ETH_P_IP) && \
73 brnf_filter_vlan_tagged)
75 #define IS_VLAN_IPV6(skb) \
76 (skb->protocol == htons(ETH_P_8021Q) && \
77 vlan_proto(skb) == htons(ETH_P_IPV6) &&\
78 brnf_filter_vlan_tagged)
80 #define IS_VLAN_ARP(skb) \
81 (skb->protocol == htons(ETH_P_8021Q) && \
82 vlan_proto(skb) == htons(ETH_P_ARP) && \
83 brnf_filter_vlan_tagged)
85 static inline __be16 pppoe_proto(const struct sk_buff *skb)
87 return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
88 sizeof(struct pppoe_hdr)));
91 #define IS_PPPOE_IP(skb) \
92 (skb->protocol == htons(ETH_P_PPP_SES) && \
93 pppoe_proto(skb) == htons(PPP_IP) && \
94 brnf_filter_pppoe_tagged)
96 #define IS_PPPOE_IPV6(skb) \
97 (skb->protocol == htons(ETH_P_PPP_SES) && \
98 pppoe_proto(skb) == htons(PPP_IPV6) && \
99 brnf_filter_pppoe_tagged)
101 static void fake_update_pmtu(struct dst_entry *dst, u32 mtu)
105 static struct dst_ops fake_dst_ops = {
107 .protocol = cpu_to_be16(ETH_P_IP),
108 .update_pmtu = fake_update_pmtu,
109 .entries = ATOMIC_INIT(0),
113 * Initialize bogus route table used to keep netfilter happy.
114 * Currently, we fill in the PMTU entry because netfilter
115 * refragmentation needs it, and the rt_flags entry because
116 * ipt_REJECT needs it. Future netfilter modules might
117 * require us to fill additional fields.
119 void br_netfilter_rtable_init(struct net_bridge *br)
121 struct rtable *rt = &br->fake_rtable;
123 atomic_set(&rt->dst.__refcnt, 1);
124 rt->dst.dev = br->dev;
125 rt->dst.path = &rt->dst;
126 rt->dst.metrics[RTAX_MTU - 1] = 1500;
127 rt->dst.flags = DST_NOXFRM;
128 rt->dst.ops = &fake_dst_ops;
131 static inline struct rtable *bridge_parent_rtable(const struct net_device *dev)
133 struct net_bridge_port *port = rcu_dereference(dev->br_port);
135 return port ? &port->br->fake_rtable : NULL;
138 static inline struct net_device *bridge_parent(const struct net_device *dev)
140 struct net_bridge_port *port = rcu_dereference(dev->br_port);
142 return port ? port->br->dev : NULL;
145 static inline struct nf_bridge_info *nf_bridge_alloc(struct sk_buff *skb)
147 skb->nf_bridge = kzalloc(sizeof(struct nf_bridge_info), GFP_ATOMIC);
148 if (likely(skb->nf_bridge))
149 atomic_set(&(skb->nf_bridge->use), 1);
151 return skb->nf_bridge;
154 static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
156 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
158 if (atomic_read(&nf_bridge->use) > 1) {
159 struct nf_bridge_info *tmp = nf_bridge_alloc(skb);
162 memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info));
163 atomic_set(&tmp->use, 1);
164 nf_bridge_put(nf_bridge);
171 static inline void nf_bridge_push_encap_header(struct sk_buff *skb)
173 unsigned int len = nf_bridge_encap_header_len(skb);
176 skb->network_header -= len;
179 static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
181 unsigned int len = nf_bridge_encap_header_len(skb);
184 skb->network_header += len;
187 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
189 unsigned int len = nf_bridge_encap_header_len(skb);
191 skb_pull_rcsum(skb, len);
192 skb->network_header += len;
195 static inline void nf_bridge_save_header(struct sk_buff *skb)
197 int header_size = ETH_HLEN + nf_bridge_encap_header_len(skb);
199 skb_copy_from_linear_data_offset(skb, -header_size,
200 skb->nf_bridge->data, header_size);
203 static inline void nf_bridge_update_protocol(struct sk_buff *skb)
205 if (skb->nf_bridge->mask & BRNF_8021Q)
206 skb->protocol = htons(ETH_P_8021Q);
207 else if (skb->nf_bridge->mask & BRNF_PPPoE)
208 skb->protocol = htons(ETH_P_PPP_SES);
211 /* Fill in the header for fragmented IP packets handled by
212 * the IPv4 connection tracking code.
214 int nf_bridge_copy_header(struct sk_buff *skb)
217 unsigned int header_size;
219 nf_bridge_update_protocol(skb);
220 header_size = ETH_HLEN + nf_bridge_encap_header_len(skb);
221 err = skb_cow_head(skb, header_size);
225 skb_copy_to_linear_data_offset(skb, -header_size,
226 skb->nf_bridge->data, header_size);
227 __skb_push(skb, nf_bridge_encap_header_len(skb));
231 /* PF_BRIDGE/PRE_ROUTING *********************************************/
232 /* Undo the changes made for ip6tables PREROUTING and continue the
233 * bridge PRE_ROUTING hook. */
234 static int br_nf_pre_routing_finish_ipv6(struct sk_buff *skb)
236 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
239 if (nf_bridge->mask & BRNF_PKT_TYPE) {
240 skb->pkt_type = PACKET_OTHERHOST;
241 nf_bridge->mask ^= BRNF_PKT_TYPE;
243 nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
245 rt = bridge_parent_rtable(nf_bridge->physindev);
250 skb_dst_set_noref(skb, &rt->dst);
252 skb->dev = nf_bridge->physindev;
253 nf_bridge_update_protocol(skb);
254 nf_bridge_push_encap_header(skb);
255 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
256 br_handle_frame_finish, 1);
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 static int br_nf_pre_routing_finish_bridge(struct sk_buff *skb)
268 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
269 struct dst_entry *dst;
271 skb->dev = bridge_parent(skb->dev);
276 neigh_hh_bridge(dst->hh, skb);
277 skb->dev = nf_bridge->physindev;
278 return br_handle_frame_finish(skb);
279 } else if (dst->neighbour) {
280 /* the neighbour function below overwrites the complete
281 * MAC header, so we save the Ethernet source address and
282 * protocol number. */
283 skb_copy_from_linear_data_offset(skb, -(ETH_HLEN-ETH_ALEN), skb->nf_bridge->data, ETH_HLEN-ETH_ALEN);
284 /* tell br_dev_xmit to continue with forwarding */
285 nf_bridge->mask |= BRNF_BRIDGED_DNAT;
286 return dst->neighbour->output(skb);
293 /* This requires some explaining. If DNAT has taken place,
294 * we will need to fix up the destination Ethernet address.
296 * There are two cases to consider:
297 * 1. The packet was DNAT'ed to a device in the same bridge
298 * port group as it was received on. We can still bridge
300 * 2. The packet was DNAT'ed to a different device, either
301 * a non-bridged device or another bridge port group.
302 * The packet will need to be routed.
304 * The correct way of distinguishing between these two cases is to
305 * call ip_route_input() and to look at skb->dst->dev, which is
306 * changed to the destination device if ip_route_input() succeeds.
308 * Let's first consider the case that ip_route_input() succeeds:
310 * If the output device equals the logical bridge device the packet
311 * came in on, we can consider this bridging. The corresponding MAC
312 * address will be obtained in br_nf_pre_routing_finish_bridge.
313 * Otherwise, the packet is considered to be routed and we just
314 * change the destination MAC address so that the packet will
315 * later be passed up to the IP stack to be routed. For a redirected
316 * packet, ip_route_input() will give back the localhost as output device,
317 * which differs from the bridge device.
319 * Let's now consider the case that ip_route_input() fails:
321 * This can be because the destination address is martian, in which case
322 * the packet will be dropped.
323 * If IP forwarding is disabled, ip_route_input() will fail, while
324 * ip_route_output_key() can return success. The source
325 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
326 * thinks we're handling a locally generated packet and won't care
327 * if IP forwarding is enabled. If the output device equals the logical bridge
328 * device, we proceed as if ip_route_input() succeeded. If it differs from the
329 * logical bridge port or if ip_route_output_key() fails we drop the packet.
331 static int br_nf_pre_routing_finish(struct sk_buff *skb)
333 struct net_device *dev = skb->dev;
334 struct iphdr *iph = ip_hdr(skb);
335 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
339 if (nf_bridge->mask & BRNF_PKT_TYPE) {
340 skb->pkt_type = PACKET_OTHERHOST;
341 nf_bridge->mask ^= BRNF_PKT_TYPE;
343 nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
344 if (dnat_took_place(skb)) {
345 if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
351 .tos = RT_TOS(iph->tos) },
355 struct in_device *in_dev = __in_dev_get_rcu(dev);
357 /* If err equals -EHOSTUNREACH the error is due to a
358 * martian destination or due to the fact that
359 * forwarding is disabled. For most martian packets,
360 * ip_route_output_key() will fail. It won't fail for 2 types of
361 * martian destinations: loopback destinations and destination
362 * 0.0.0.0. In both cases the packet will be dropped because the
363 * destination is the loopback device and not the bridge. */
364 if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
367 if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
368 /* - Bridged-and-DNAT'ed traffic doesn't
369 * require ip_forwarding. */
370 if (((struct dst_entry *)rt)->dev == dev) {
371 skb_dst_set(skb, (struct dst_entry *)rt);
374 dst_release((struct dst_entry *)rt);
380 if (skb_dst(skb)->dev == dev) {
382 skb->dev = nf_bridge->physindev;
383 nf_bridge_update_protocol(skb);
384 nf_bridge_push_encap_header(skb);
385 NF_HOOK_THRESH(NFPROTO_BRIDGE,
388 br_nf_pre_routing_finish_bridge,
392 memcpy(eth_hdr(skb)->h_dest, dev->dev_addr, ETH_ALEN);
393 skb->pkt_type = PACKET_HOST;
396 rt = bridge_parent_rtable(nf_bridge->physindev);
401 skb_dst_set_noref(skb, &rt->dst);
404 skb->dev = nf_bridge->physindev;
405 nf_bridge_update_protocol(skb);
406 nf_bridge_push_encap_header(skb);
407 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
408 br_handle_frame_finish, 1);
413 /* Some common code for IPv4/IPv6 */
414 static struct net_device *setup_pre_routing(struct sk_buff *skb)
416 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
418 if (skb->pkt_type == PACKET_OTHERHOST) {
419 skb->pkt_type = PACKET_HOST;
420 nf_bridge->mask |= BRNF_PKT_TYPE;
423 nf_bridge->mask |= BRNF_NF_BRIDGE_PREROUTING;
424 nf_bridge->physindev = skb->dev;
425 skb->dev = bridge_parent(skb->dev);
426 if (skb->protocol == htons(ETH_P_8021Q))
427 nf_bridge->mask |= BRNF_8021Q;
428 else if (skb->protocol == htons(ETH_P_PPP_SES))
429 nf_bridge->mask |= BRNF_PPPoE;
434 /* We only check the length. A bridge shouldn't do any hop-by-hop stuff anyway */
435 static int check_hbh_len(struct sk_buff *skb)
437 unsigned char *raw = (u8 *)(ipv6_hdr(skb) + 1);
439 const unsigned char *nh = skb_network_header(skb);
441 int len = (raw[1] + 1) << 3;
443 if ((raw + len) - skb->data > skb_headlen(skb))
450 int optlen = nh[off + 1] + 2;
461 if (nh[off + 1] != 4 || (off & 3) != 2)
463 pkt_len = ntohl(*(__be32 *) (nh + off + 2));
464 if (pkt_len <= IPV6_MAXPLEN ||
465 ipv6_hdr(skb)->payload_len)
467 if (pkt_len > skb->len - sizeof(struct ipv6hdr))
469 if (pskb_trim_rcsum(skb,
470 pkt_len + sizeof(struct ipv6hdr)))
472 nh = skb_network_header(skb);
489 /* Replicate the checks that IPv6 does on packet reception and pass the packet
490 * to ip6tables, which doesn't support NAT, so things are fairly simple. */
491 static unsigned int br_nf_pre_routing_ipv6(unsigned int hook,
493 const struct net_device *in,
494 const struct net_device *out,
495 int (*okfn)(struct sk_buff *))
500 if (skb->len < sizeof(struct ipv6hdr))
503 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
508 if (hdr->version != 6)
511 pkt_len = ntohs(hdr->payload_len);
513 if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
514 if (pkt_len + sizeof(struct ipv6hdr) > skb->len)
516 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
519 if (hdr->nexthdr == NEXTHDR_HOP && check_hbh_len(skb))
522 nf_bridge_put(skb->nf_bridge);
523 if (!nf_bridge_alloc(skb))
525 if (!setup_pre_routing(skb))
528 skb->protocol = htons(ETH_P_IPV6);
529 NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
530 br_nf_pre_routing_finish_ipv6);
538 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
539 * Replicate the checks that IPv4 does on packet reception.
540 * Set skb->dev to the bridge device (i.e. parent of the
541 * receiving device) to make netfilter happy, the REDIRECT
542 * target in particular. Save the original destination IP
543 * address to be able to detect DNAT afterwards. */
544 static unsigned int br_nf_pre_routing(unsigned int hook, struct sk_buff *skb,
545 const struct net_device *in,
546 const struct net_device *out,
547 int (*okfn)(struct sk_buff *))
549 struct net_bridge_port *p;
550 struct net_bridge *br;
552 __u32 len = nf_bridge_encap_header_len(skb);
554 if (unlikely(!pskb_may_pull(skb, len)))
557 p = rcu_dereference(in->br_port);
562 if (skb->protocol == htons(ETH_P_IPV6) || IS_VLAN_IPV6(skb) ||
563 IS_PPPOE_IPV6(skb)) {
564 if (!brnf_call_ip6tables && !br->nf_call_ip6tables)
567 nf_bridge_pull_encap_header_rcsum(skb);
568 return br_nf_pre_routing_ipv6(hook, skb, in, out, okfn);
571 if (!brnf_call_iptables && !br->nf_call_iptables)
574 if (skb->protocol != htons(ETH_P_IP) && !IS_VLAN_IP(skb) &&
578 nf_bridge_pull_encap_header_rcsum(skb);
580 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
584 if (iph->ihl < 5 || iph->version != 4)
587 if (!pskb_may_pull(skb, 4 * iph->ihl))
591 if (ip_fast_csum((__u8 *) iph, iph->ihl) != 0)
594 len = ntohs(iph->tot_len);
595 if (skb->len < len || len < 4 * iph->ihl)
598 pskb_trim_rcsum(skb, len);
600 nf_bridge_put(skb->nf_bridge);
601 if (!nf_bridge_alloc(skb))
603 if (!setup_pre_routing(skb))
605 store_orig_dstaddr(skb);
606 skb->protocol = htons(ETH_P_IP);
608 NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
609 br_nf_pre_routing_finish);
614 // IP_INC_STATS_BH(IpInHdrErrors);
620 /* PF_BRIDGE/LOCAL_IN ************************************************/
621 /* The packet is locally destined, which requires a real
622 * dst_entry, so detach the fake one. On the way up, the
623 * packet would pass through PRE_ROUTING again (which already
624 * took place when the packet entered the bridge), but we
625 * register an IPv4 PRE_ROUTING 'sabotage' hook that will
626 * prevent this from happening. */
627 static unsigned int br_nf_local_in(unsigned int hook, struct sk_buff *skb,
628 const struct net_device *in,
629 const struct net_device *out,
630 int (*okfn)(struct sk_buff *))
632 struct rtable *rt = skb_rtable(skb);
634 if (rt && rt == bridge_parent_rtable(in))
640 /* PF_BRIDGE/FORWARD *************************************************/
641 static int br_nf_forward_finish(struct sk_buff *skb)
643 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
644 struct net_device *in;
646 if (skb->protocol != htons(ETH_P_ARP) && !IS_VLAN_ARP(skb)) {
647 in = nf_bridge->physindev;
648 if (nf_bridge->mask & BRNF_PKT_TYPE) {
649 skb->pkt_type = PACKET_OTHERHOST;
650 nf_bridge->mask ^= BRNF_PKT_TYPE;
652 nf_bridge_update_protocol(skb);
654 in = *((struct net_device **)(skb->cb));
656 nf_bridge_push_encap_header(skb);
658 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_FORWARD, skb, in,
659 skb->dev, br_forward_finish, 1);
663 /* This is the 'purely bridged' case. For IP, we pass the packet to
664 * netfilter with indev and outdev set to the bridge device,
665 * but we are still able to filter on the 'real' indev/outdev
666 * because of the physdev module. For ARP, indev and outdev are the
668 static unsigned int br_nf_forward_ip(unsigned int hook, struct sk_buff *skb,
669 const struct net_device *in,
670 const struct net_device *out,
671 int (*okfn)(struct sk_buff *))
673 struct nf_bridge_info *nf_bridge;
674 struct net_device *parent;
680 /* Need exclusive nf_bridge_info since we might have multiple
681 * different physoutdevs. */
682 if (!nf_bridge_unshare(skb))
685 parent = bridge_parent(out);
689 if (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb) ||
692 else if (skb->protocol == htons(ETH_P_IPV6) || IS_VLAN_IPV6(skb) ||
698 nf_bridge_pull_encap_header(skb);
700 nf_bridge = skb->nf_bridge;
701 if (skb->pkt_type == PACKET_OTHERHOST) {
702 skb->pkt_type = PACKET_HOST;
703 nf_bridge->mask |= BRNF_PKT_TYPE;
706 /* The physdev module checks on this */
707 nf_bridge->mask |= BRNF_BRIDGED;
708 nf_bridge->physoutdev = skb->dev;
710 skb->protocol = htons(ETH_P_IP);
712 skb->protocol = htons(ETH_P_IPV6);
714 NF_HOOK(pf, NF_INET_FORWARD, skb, bridge_parent(in), parent,
715 br_nf_forward_finish);
720 static unsigned int br_nf_forward_arp(unsigned int hook, struct sk_buff *skb,
721 const struct net_device *in,
722 const struct net_device *out,
723 int (*okfn)(struct sk_buff *))
725 struct net_bridge_port *p;
726 struct net_bridge *br;
727 struct net_device **d = (struct net_device **)(skb->cb);
729 p = rcu_dereference(out->br_port);
734 if (!brnf_call_arptables && !br->nf_call_arptables)
737 if (skb->protocol != htons(ETH_P_ARP)) {
738 if (!IS_VLAN_ARP(skb))
740 nf_bridge_pull_encap_header(skb);
743 if (arp_hdr(skb)->ar_pln != 4) {
744 if (IS_VLAN_ARP(skb))
745 nf_bridge_push_encap_header(skb);
748 *d = (struct net_device *)in;
749 NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, skb, (struct net_device *)in,
750 (struct net_device *)out, br_nf_forward_finish);
755 #if defined(CONFIG_NF_CONNTRACK_IPV4) || defined(CONFIG_NF_CONNTRACK_IPV4_MODULE)
756 static int br_nf_dev_queue_xmit(struct sk_buff *skb)
758 if (skb->nfct != NULL && skb->protocol == htons(ETH_P_IP) &&
759 skb->len + nf_bridge_mtu_reduction(skb) > skb->dev->mtu &&
761 return ip_fragment(skb, br_dev_queue_push_xmit);
763 return br_dev_queue_push_xmit(skb);
766 static int br_nf_dev_queue_xmit(struct sk_buff *skb)
768 return br_dev_queue_push_xmit(skb);
772 /* PF_BRIDGE/POST_ROUTING ********************************************/
773 static unsigned int br_nf_post_routing(unsigned int hook, struct sk_buff *skb,
774 const struct net_device *in,
775 const struct net_device *out,
776 int (*okfn)(struct sk_buff *))
778 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
779 struct net_device *realoutdev = bridge_parent(skb->dev);
782 if (!nf_bridge || !(nf_bridge->mask & BRNF_BRIDGED))
788 if (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb) ||
791 else if (skb->protocol == htons(ETH_P_IPV6) || IS_VLAN_IPV6(skb) ||
797 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
798 * about the value of skb->pkt_type. */
799 if (skb->pkt_type == PACKET_OTHERHOST) {
800 skb->pkt_type = PACKET_HOST;
801 nf_bridge->mask |= BRNF_PKT_TYPE;
804 nf_bridge_pull_encap_header(skb);
805 nf_bridge_save_header(skb);
807 skb->protocol = htons(ETH_P_IP);
809 skb->protocol = htons(ETH_P_IPV6);
811 NF_HOOK(pf, NF_INET_POST_ROUTING, skb, NULL, realoutdev,
812 br_nf_dev_queue_xmit);
817 /* IP/SABOTAGE *****************************************************/
818 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
819 * for the second time. */
820 static unsigned int ip_sabotage_in(unsigned int hook, struct sk_buff *skb,
821 const struct net_device *in,
822 const struct net_device *out,
823 int (*okfn)(struct sk_buff *))
825 if (skb->nf_bridge &&
826 !(skb->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)) {
833 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
834 * br_dev_queue_push_xmit is called afterwards */
835 static struct nf_hook_ops br_nf_ops[] __read_mostly = {
837 .hook = br_nf_pre_routing,
838 .owner = THIS_MODULE,
840 .hooknum = NF_BR_PRE_ROUTING,
841 .priority = NF_BR_PRI_BRNF,
844 .hook = br_nf_local_in,
845 .owner = THIS_MODULE,
847 .hooknum = NF_BR_LOCAL_IN,
848 .priority = NF_BR_PRI_BRNF,
851 .hook = br_nf_forward_ip,
852 .owner = THIS_MODULE,
854 .hooknum = NF_BR_FORWARD,
855 .priority = NF_BR_PRI_BRNF - 1,
858 .hook = br_nf_forward_arp,
859 .owner = THIS_MODULE,
861 .hooknum = NF_BR_FORWARD,
862 .priority = NF_BR_PRI_BRNF,
865 .hook = br_nf_post_routing,
866 .owner = THIS_MODULE,
868 .hooknum = NF_BR_POST_ROUTING,
869 .priority = NF_BR_PRI_LAST,
872 .hook = ip_sabotage_in,
873 .owner = THIS_MODULE,
875 .hooknum = NF_INET_PRE_ROUTING,
876 .priority = NF_IP_PRI_FIRST,
879 .hook = ip_sabotage_in,
880 .owner = THIS_MODULE,
882 .hooknum = NF_INET_PRE_ROUTING,
883 .priority = NF_IP6_PRI_FIRST,
889 int brnf_sysctl_call_tables(ctl_table * ctl, int write,
890 void __user * buffer, size_t * lenp, loff_t * ppos)
894 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
896 if (write && *(int *)(ctl->data))
897 *(int *)(ctl->data) = 1;
901 static ctl_table brnf_table[] = {
903 .procname = "bridge-nf-call-arptables",
904 .data = &brnf_call_arptables,
905 .maxlen = sizeof(int),
907 .proc_handler = brnf_sysctl_call_tables,
910 .procname = "bridge-nf-call-iptables",
911 .data = &brnf_call_iptables,
912 .maxlen = sizeof(int),
914 .proc_handler = brnf_sysctl_call_tables,
917 .procname = "bridge-nf-call-ip6tables",
918 .data = &brnf_call_ip6tables,
919 .maxlen = sizeof(int),
921 .proc_handler = brnf_sysctl_call_tables,
924 .procname = "bridge-nf-filter-vlan-tagged",
925 .data = &brnf_filter_vlan_tagged,
926 .maxlen = sizeof(int),
928 .proc_handler = brnf_sysctl_call_tables,
931 .procname = "bridge-nf-filter-pppoe-tagged",
932 .data = &brnf_filter_pppoe_tagged,
933 .maxlen = sizeof(int),
935 .proc_handler = brnf_sysctl_call_tables,
940 static struct ctl_path brnf_path[] = {
941 { .procname = "net", },
942 { .procname = "bridge", },
947 int __init br_netfilter_init(void)
951 ret = nf_register_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
955 brnf_sysctl_header = register_sysctl_paths(brnf_path, brnf_table);
956 if (brnf_sysctl_header == NULL) {
958 "br_netfilter: can't register to sysctl.\n");
959 nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
963 printk(KERN_NOTICE "Bridge firewalling registered\n");
967 void br_netfilter_fini(void)
969 nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
971 unregister_sysctl_table(brnf_sysctl_header);