3 * Linux ethernet bridge
6 * Lennert Buytenhek <buytenh@gnu.org>
7 * Bart De Schuymer (maintainer) <bdschuym@pandora.be>
10 * Apr 29 2003: physdev module support (bdschuym)
11 * Jun 19 2003: let arptables see bridged ARP traffic (bdschuym)
12 * Oct 06 2003: filter encapsulated IP/ARP VLAN traffic on untagged bridge
14 * Sep 01 2004: add IPv6 filtering (bdschuym)
16 * This program is free software; you can redistribute it and/or
17 * modify it under the terms of the GNU General Public License
18 * as published by the Free Software Foundation; either version
19 * 2 of the License, or (at your option) any later version.
21 * Lennert dedicates this file to Kerstin Wurdinger.
24 #include <linux/module.h>
25 #include <linux/kernel.h>
27 #include <linux/netdevice.h>
28 #include <linux/skbuff.h>
29 #include <linux/if_arp.h>
30 #include <linux/if_ether.h>
31 #include <linux/if_vlan.h>
32 #include <linux/if_pppox.h>
33 #include <linux/ppp_defs.h>
34 #include <linux/netfilter_bridge.h>
35 #include <linux/netfilter_ipv4.h>
36 #include <linux/netfilter_ipv6.h>
37 #include <linux/netfilter_arp.h>
38 #include <linux/in_route.h>
39 #include <linux/inetdevice.h>
43 #include <net/route.h>
45 #include <asm/uaccess.h>
46 #include "br_private.h"
48 #include <linux/sysctl.h>
51 #define skb_origaddr(skb) (((struct bridge_skb_cb *) \
52 (skb->nf_bridge->data))->daddr.ipv4)
53 #define store_orig_dstaddr(skb) (skb_origaddr(skb) = ip_hdr(skb)->daddr)
54 #define dnat_took_place(skb) (skb_origaddr(skb) != ip_hdr(skb)->daddr)
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 = 1;
62 static int brnf_filter_pppoe_tagged __read_mostly = 1;
64 #define brnf_filter_vlan_tagged 1
65 #define brnf_filter_pppoe_tagged 1
68 static inline __be16 vlan_proto(const struct sk_buff *skb)
70 return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
73 #define IS_VLAN_IP(skb) \
74 (skb->protocol == htons(ETH_P_8021Q) && \
75 vlan_proto(skb) == htons(ETH_P_IP) && \
76 brnf_filter_vlan_tagged)
78 #define IS_VLAN_IPV6(skb) \
79 (skb->protocol == htons(ETH_P_8021Q) && \
80 vlan_proto(skb) == htons(ETH_P_IPV6) &&\
81 brnf_filter_vlan_tagged)
83 #define IS_VLAN_ARP(skb) \
84 (skb->protocol == htons(ETH_P_8021Q) && \
85 vlan_proto(skb) == htons(ETH_P_ARP) && \
86 brnf_filter_vlan_tagged)
88 static inline __be16 pppoe_proto(const struct sk_buff *skb)
90 return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
91 sizeof(struct pppoe_hdr)));
94 #define IS_PPPOE_IP(skb) \
95 (skb->protocol == htons(ETH_P_PPP_SES) && \
96 pppoe_proto(skb) == htons(PPP_IP) && \
97 brnf_filter_pppoe_tagged)
99 #define IS_PPPOE_IPV6(skb) \
100 (skb->protocol == htons(ETH_P_PPP_SES) && \
101 pppoe_proto(skb) == htons(PPP_IPV6) && \
102 brnf_filter_pppoe_tagged)
104 /* We need these fake structures to make netfilter happy --
105 * lots of places assume that skb->dst != NULL, which isn't
106 * all that unreasonable.
108 * Currently, we fill in the PMTU entry because netfilter
109 * refragmentation needs it, and the rt_flags entry because
110 * ipt_REJECT needs it. Future netfilter modules might
111 * require us to fill additional fields. */
112 static struct net_device __fake_net_device = {
113 .hard_header_len = ETH_HLEN
116 static struct rtable __fake_rtable = {
119 .__refcnt = ATOMIC_INIT(1),
120 .dev = &__fake_net_device,
121 .path = &__fake_rtable.u.dst,
122 .metrics = {[RTAX_MTU - 1] = 1500},
129 static inline struct net_device *bridge_parent(const struct net_device *dev)
131 struct net_bridge_port *port = rcu_dereference(dev->br_port);
133 return port ? port->br->dev : NULL;
136 static inline struct nf_bridge_info *nf_bridge_alloc(struct sk_buff *skb)
138 skb->nf_bridge = kzalloc(sizeof(struct nf_bridge_info), GFP_ATOMIC);
139 if (likely(skb->nf_bridge))
140 atomic_set(&(skb->nf_bridge->use), 1);
142 return skb->nf_bridge;
145 static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
147 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
149 if (atomic_read(&nf_bridge->use) > 1) {
150 struct nf_bridge_info *tmp = nf_bridge_alloc(skb);
153 memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info));
154 atomic_set(&tmp->use, 1);
155 nf_bridge_put(nf_bridge);
162 static inline void nf_bridge_push_encap_header(struct sk_buff *skb)
164 unsigned int len = nf_bridge_encap_header_len(skb);
167 skb->network_header -= len;
170 static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
172 unsigned int len = nf_bridge_encap_header_len(skb);
175 skb->network_header += len;
178 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
180 unsigned int len = nf_bridge_encap_header_len(skb);
182 skb_pull_rcsum(skb, len);
183 skb->network_header += len;
186 static inline void nf_bridge_save_header(struct sk_buff *skb)
188 int header_size = ETH_HLEN + nf_bridge_encap_header_len(skb);
190 skb_copy_from_linear_data_offset(skb, -header_size,
191 skb->nf_bridge->data, header_size);
195 * When forwarding bridge frames, we save a copy of the original
196 * header before processing.
198 int nf_bridge_copy_header(struct sk_buff *skb)
201 int header_size = ETH_HLEN + nf_bridge_encap_header_len(skb);
203 err = skb_cow_head(skb, header_size);
207 skb_copy_to_linear_data_offset(skb, -header_size,
208 skb->nf_bridge->data, header_size);
209 __skb_push(skb, nf_bridge_encap_header_len(skb));
213 /* PF_BRIDGE/PRE_ROUTING *********************************************/
214 /* Undo the changes made for ip6tables PREROUTING and continue the
215 * bridge PRE_ROUTING hook. */
216 static int br_nf_pre_routing_finish_ipv6(struct sk_buff *skb)
218 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
220 if (nf_bridge->mask & BRNF_PKT_TYPE) {
221 skb->pkt_type = PACKET_OTHERHOST;
222 nf_bridge->mask ^= BRNF_PKT_TYPE;
224 nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
226 skb->dst = (struct dst_entry *)&__fake_rtable;
229 skb->dev = nf_bridge->physindev;
230 nf_bridge_push_encap_header(skb);
231 NF_HOOK_THRESH(PF_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
232 br_handle_frame_finish, 1);
237 static void __br_dnat_complain(void)
239 static unsigned long last_complaint;
241 if (jiffies - last_complaint >= 5 * HZ) {
242 printk(KERN_WARNING "Performing cross-bridge DNAT requires IP "
243 "forwarding to be enabled\n");
244 last_complaint = jiffies;
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;
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;
327 if (nf_bridge->mask & BRNF_PKT_TYPE) {
328 skb->pkt_type = PACKET_OTHERHOST;
329 nf_bridge->mask ^= BRNF_PKT_TYPE;
331 nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
332 if (dnat_took_place(skb)) {
333 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(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(&init_net, &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 = (struct dst_entry *)rt;
363 /* we are sure that forwarding is disabled, so printing
364 * this message is no problem. Note that the packet could
365 * still have a martian destination address, in which case
366 * the packet could be dropped even if forwarding were enabled */
367 __br_dnat_complain();
368 dst_release((struct dst_entry *)rt);
374 if (skb->dst->dev == dev) {
376 /* Tell br_nf_local_out this is a
378 nf_bridge->mask |= BRNF_BRIDGED_DNAT;
379 skb->dev = nf_bridge->physindev;
380 nf_bridge_push_encap_header(skb);
381 NF_HOOK_THRESH(PF_BRIDGE, NF_BR_PRE_ROUTING,
383 br_nf_pre_routing_finish_bridge,
387 memcpy(eth_hdr(skb)->h_dest, dev->dev_addr, ETH_ALEN);
388 skb->pkt_type = PACKET_HOST;
391 skb->dst = (struct dst_entry *)&__fake_rtable;
395 skb->dev = nf_bridge->physindev;
396 nf_bridge_push_encap_header(skb);
397 NF_HOOK_THRESH(PF_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
398 br_handle_frame_finish, 1);
403 /* Some common code for IPv4/IPv6 */
404 static struct net_device *setup_pre_routing(struct sk_buff *skb)
406 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
408 if (skb->pkt_type == PACKET_OTHERHOST) {
409 skb->pkt_type = PACKET_HOST;
410 nf_bridge->mask |= BRNF_PKT_TYPE;
413 nf_bridge->mask |= BRNF_NF_BRIDGE_PREROUTING;
414 nf_bridge->physindev = skb->dev;
415 skb->dev = bridge_parent(skb->dev);
420 /* We only check the length. A bridge shouldn't do any hop-by-hop stuff anyway */
421 static int check_hbh_len(struct sk_buff *skb)
423 unsigned char *raw = (u8 *)(ipv6_hdr(skb) + 1);
425 const unsigned char *nh = skb_network_header(skb);
427 int len = (raw[1] + 1) << 3;
429 if ((raw + len) - skb->data > skb_headlen(skb))
436 int optlen = nh[off + 1] + 2;
447 if (nh[off + 1] != 4 || (off & 3) != 2)
449 pkt_len = ntohl(*(__be32 *) (nh + off + 2));
450 if (pkt_len <= IPV6_MAXPLEN ||
451 ipv6_hdr(skb)->payload_len)
453 if (pkt_len > skb->len - sizeof(struct ipv6hdr))
455 if (pskb_trim_rcsum(skb,
456 pkt_len + sizeof(struct ipv6hdr)))
458 nh = skb_network_header(skb);
475 /* Replicate the checks that IPv6 does on packet reception and pass the packet
476 * to ip6tables, which doesn't support NAT, so things are fairly simple. */
477 static unsigned int br_nf_pre_routing_ipv6(unsigned int hook,
479 const struct net_device *in,
480 const struct net_device *out,
481 int (*okfn)(struct sk_buff *))
486 if (skb->len < sizeof(struct ipv6hdr))
489 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
494 if (hdr->version != 6)
497 pkt_len = ntohs(hdr->payload_len);
499 if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
500 if (pkt_len + sizeof(struct ipv6hdr) > skb->len)
502 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
505 if (hdr->nexthdr == NEXTHDR_HOP && check_hbh_len(skb))
508 nf_bridge_put(skb->nf_bridge);
509 if (!nf_bridge_alloc(skb))
511 if (!setup_pre_routing(skb))
514 NF_HOOK(PF_INET6, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
515 br_nf_pre_routing_finish_ipv6);
523 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
524 * Replicate the checks that IPv4 does on packet reception.
525 * Set skb->dev to the bridge device (i.e. parent of the
526 * receiving device) to make netfilter happy, the REDIRECT
527 * target in particular. Save the original destination IP
528 * address to be able to detect DNAT afterwards. */
529 static unsigned int br_nf_pre_routing(unsigned int hook, struct sk_buff *skb,
530 const struct net_device *in,
531 const struct net_device *out,
532 int (*okfn)(struct sk_buff *))
535 __u32 len = nf_bridge_encap_header_len(skb);
537 if (unlikely(!pskb_may_pull(skb, len)))
540 if (skb->protocol == htons(ETH_P_IPV6) || IS_VLAN_IPV6(skb) ||
541 IS_PPPOE_IPV6(skb)) {
543 if (!brnf_call_ip6tables)
546 nf_bridge_pull_encap_header_rcsum(skb);
547 return br_nf_pre_routing_ipv6(hook, skb, in, out, okfn);
550 if (!brnf_call_iptables)
554 if (skb->protocol != htons(ETH_P_IP) && !IS_VLAN_IP(skb) &&
558 nf_bridge_pull_encap_header_rcsum(skb);
560 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
564 if (iph->ihl < 5 || iph->version != 4)
567 if (!pskb_may_pull(skb, 4 * iph->ihl))
571 if (ip_fast_csum((__u8 *) iph, iph->ihl) != 0)
574 len = ntohs(iph->tot_len);
575 if (skb->len < len || len < 4 * iph->ihl)
578 pskb_trim_rcsum(skb, len);
580 nf_bridge_put(skb->nf_bridge);
581 if (!nf_bridge_alloc(skb))
583 if (!setup_pre_routing(skb))
585 store_orig_dstaddr(skb);
587 NF_HOOK(PF_INET, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
588 br_nf_pre_routing_finish);
593 // IP_INC_STATS_BH(IpInHdrErrors);
599 /* PF_BRIDGE/LOCAL_IN ************************************************/
600 /* The packet is locally destined, which requires a real
601 * dst_entry, so detach the fake one. On the way up, the
602 * packet would pass through PRE_ROUTING again (which already
603 * took place when the packet entered the bridge), but we
604 * register an IPv4 PRE_ROUTING 'sabotage' hook that will
605 * prevent this from happening. */
606 static unsigned int br_nf_local_in(unsigned int hook, struct sk_buff *skb,
607 const struct net_device *in,
608 const struct net_device *out,
609 int (*okfn)(struct sk_buff *))
611 if (skb->dst == (struct dst_entry *)&__fake_rtable) {
612 dst_release(skb->dst);
619 /* PF_BRIDGE/FORWARD *************************************************/
620 static int br_nf_forward_finish(struct sk_buff *skb)
622 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
623 struct net_device *in;
625 if (skb->protocol != htons(ETH_P_ARP) && !IS_VLAN_ARP(skb)) {
626 in = nf_bridge->physindev;
627 if (nf_bridge->mask & BRNF_PKT_TYPE) {
628 skb->pkt_type = PACKET_OTHERHOST;
629 nf_bridge->mask ^= BRNF_PKT_TYPE;
632 in = *((struct net_device **)(skb->cb));
634 nf_bridge_push_encap_header(skb);
635 NF_HOOK_THRESH(PF_BRIDGE, NF_BR_FORWARD, skb, in,
636 skb->dev, br_forward_finish, 1);
640 /* This is the 'purely bridged' case. For IP, we pass the packet to
641 * netfilter with indev and outdev set to the bridge device,
642 * but we are still able to filter on the 'real' indev/outdev
643 * because of the physdev module. For ARP, indev and outdev are the
645 static unsigned int br_nf_forward_ip(unsigned int hook, struct sk_buff *skb,
646 const struct net_device *in,
647 const struct net_device *out,
648 int (*okfn)(struct sk_buff *))
650 struct nf_bridge_info *nf_bridge;
651 struct net_device *parent;
657 /* Need exclusive nf_bridge_info since we might have multiple
658 * different physoutdevs. */
659 if (!nf_bridge_unshare(skb))
662 parent = bridge_parent(out);
666 if (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb) ||
672 nf_bridge_pull_encap_header(skb);
674 nf_bridge = skb->nf_bridge;
675 if (skb->pkt_type == PACKET_OTHERHOST) {
676 skb->pkt_type = PACKET_HOST;
677 nf_bridge->mask |= BRNF_PKT_TYPE;
680 /* The physdev module checks on this */
681 nf_bridge->mask |= BRNF_BRIDGED;
682 nf_bridge->physoutdev = skb->dev;
684 NF_HOOK(pf, NF_INET_FORWARD, skb, bridge_parent(in), parent,
685 br_nf_forward_finish);
690 static unsigned int br_nf_forward_arp(unsigned int hook, struct sk_buff *skb,
691 const struct net_device *in,
692 const struct net_device *out,
693 int (*okfn)(struct sk_buff *))
695 struct net_device **d = (struct net_device **)(skb->cb);
698 if (!brnf_call_arptables)
702 if (skb->protocol != htons(ETH_P_ARP)) {
703 if (!IS_VLAN_ARP(skb))
705 nf_bridge_pull_encap_header(skb);
708 if (arp_hdr(skb)->ar_pln != 4) {
709 if (IS_VLAN_ARP(skb))
710 nf_bridge_push_encap_header(skb);
713 *d = (struct net_device *)in;
714 NF_HOOK(NF_ARP, NF_ARP_FORWARD, skb, (struct net_device *)in,
715 (struct net_device *)out, br_nf_forward_finish);
720 /* PF_BRIDGE/LOCAL_OUT ***********************************************
722 * This function sees both locally originated IP packets and forwarded
723 * IP packets (in both cases the destination device is a bridge
724 * device). It also sees bridged-and-DNAT'ed packets.
726 * If (nf_bridge->mask & BRNF_BRIDGED_DNAT) then the packet is bridged
727 * and we fake the PF_BRIDGE/FORWARD hook. The function br_nf_forward()
728 * will then fake the PF_INET/FORWARD hook. br_nf_local_out() has priority
729 * NF_BR_PRI_FIRST, so no relevant PF_BRIDGE/INPUT functions have been nor
732 static unsigned int br_nf_local_out(unsigned int hook, struct sk_buff *skb,
733 const struct net_device *in,
734 const struct net_device *out,
735 int (*okfn)(struct sk_buff *))
737 struct net_device *realindev;
738 struct nf_bridge_info *nf_bridge;
743 /* Need exclusive nf_bridge_info since we might have multiple
744 * different physoutdevs. */
745 if (!nf_bridge_unshare(skb))
748 nf_bridge = skb->nf_bridge;
749 if (!(nf_bridge->mask & BRNF_BRIDGED_DNAT))
752 /* Bridged, take PF_BRIDGE/FORWARD.
753 * (see big note in front of br_nf_pre_routing_finish) */
754 nf_bridge->physoutdev = skb->dev;
755 realindev = nf_bridge->physindev;
757 if (nf_bridge->mask & BRNF_PKT_TYPE) {
758 skb->pkt_type = PACKET_OTHERHOST;
759 nf_bridge->mask ^= BRNF_PKT_TYPE;
761 nf_bridge_push_encap_header(skb);
763 NF_HOOK(PF_BRIDGE, NF_BR_FORWARD, skb, realindev, skb->dev,
768 static int br_nf_dev_queue_xmit(struct sk_buff *skb)
770 if (skb->protocol == htons(ETH_P_IP) &&
771 skb->len > skb->dev->mtu &&
773 return ip_fragment(skb, br_dev_queue_push_xmit);
775 return br_dev_queue_push_xmit(skb);
778 /* PF_BRIDGE/POST_ROUTING ********************************************/
779 static unsigned int br_nf_post_routing(unsigned int hook, struct sk_buff *skb,
780 const struct net_device *in,
781 const struct net_device *out,
782 int (*okfn)(struct sk_buff *))
784 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
785 struct net_device *realoutdev = bridge_parent(skb->dev);
788 #ifdef CONFIG_NETFILTER_DEBUG
789 /* Be very paranoid. This probably won't happen anymore, but let's
790 * keep the check just to be sure... */
791 if (skb_mac_header(skb) < skb->head ||
792 skb_mac_header(skb) + ETH_HLEN > skb->data) {
793 printk(KERN_CRIT "br_netfilter: Argh!! br_nf_post_routing: "
794 "bad mac.raw pointer.\n");
802 if (!(nf_bridge->mask & (BRNF_BRIDGED | BRNF_BRIDGED_DNAT)))
808 if (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb) ||
814 #ifdef CONFIG_NETFILTER_DEBUG
815 if (skb->dst == NULL) {
816 printk(KERN_INFO "br_netfilter post_routing: skb->dst == NULL\n");
821 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
822 * about the value of skb->pkt_type. */
823 if (skb->pkt_type == PACKET_OTHERHOST) {
824 skb->pkt_type = PACKET_HOST;
825 nf_bridge->mask |= BRNF_PKT_TYPE;
828 nf_bridge_pull_encap_header(skb);
829 nf_bridge_save_header(skb);
831 NF_HOOK(pf, NF_INET_POST_ROUTING, skb, NULL, realoutdev,
832 br_nf_dev_queue_xmit);
836 #ifdef CONFIG_NETFILTER_DEBUG
838 if (skb->dev != NULL) {
839 printk("[%s]", skb->dev->name);
841 printk("[%s]", realoutdev->name);
843 printk(" head:%p, raw:%p, data:%p\n", skb->head, skb_mac_header(skb),
850 /* IP/SABOTAGE *****************************************************/
851 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
852 * for the second time. */
853 static unsigned int ip_sabotage_in(unsigned int hook, struct sk_buff *skb,
854 const struct net_device *in,
855 const struct net_device *out,
856 int (*okfn)(struct sk_buff *))
858 if (skb->nf_bridge &&
859 !(skb->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)) {
866 /* For br_nf_local_out we need (prio = NF_BR_PRI_FIRST), to insure that innocent
867 * PF_BRIDGE/NF_BR_LOCAL_OUT functions don't get bridged traffic as input.
868 * For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
869 * ip_refrag() can return NF_STOLEN. */
870 static struct nf_hook_ops br_nf_ops[] __read_mostly = {
871 { .hook = br_nf_pre_routing,
872 .owner = THIS_MODULE,
874 .hooknum = NF_BR_PRE_ROUTING,
875 .priority = NF_BR_PRI_BRNF, },
876 { .hook = br_nf_local_in,
877 .owner = THIS_MODULE,
879 .hooknum = NF_BR_LOCAL_IN,
880 .priority = NF_BR_PRI_BRNF, },
881 { .hook = br_nf_forward_ip,
882 .owner = THIS_MODULE,
884 .hooknum = NF_BR_FORWARD,
885 .priority = NF_BR_PRI_BRNF - 1, },
886 { .hook = br_nf_forward_arp,
887 .owner = THIS_MODULE,
889 .hooknum = NF_BR_FORWARD,
890 .priority = NF_BR_PRI_BRNF, },
891 { .hook = br_nf_local_out,
892 .owner = THIS_MODULE,
894 .hooknum = NF_BR_LOCAL_OUT,
895 .priority = NF_BR_PRI_FIRST, },
896 { .hook = br_nf_post_routing,
897 .owner = THIS_MODULE,
899 .hooknum = NF_BR_POST_ROUTING,
900 .priority = NF_BR_PRI_LAST, },
901 { .hook = ip_sabotage_in,
902 .owner = THIS_MODULE,
904 .hooknum = NF_INET_PRE_ROUTING,
905 .priority = NF_IP_PRI_FIRST, },
906 { .hook = ip_sabotage_in,
907 .owner = THIS_MODULE,
909 .hooknum = NF_INET_PRE_ROUTING,
910 .priority = NF_IP6_PRI_FIRST, },
915 int brnf_sysctl_call_tables(ctl_table * ctl, int write, struct file *filp,
916 void __user * buffer, size_t * lenp, loff_t * ppos)
920 ret = proc_dointvec(ctl, write, filp, 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", .ctl_name = CTL_NET, },
968 { .procname = "bridge", .ctl_name = NET_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);