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
,
117 static struct rtable __fake_rtable
= {
120 .__refcnt
= ATOMIC_INIT(1),
121 .dev
= &__fake_net_device
,
122 .path
= &__fake_rtable
.u
.dst
,
123 .metrics
= {[RTAX_MTU
- 1] = 1500},
130 static inline struct net_device
*bridge_parent(const struct net_device
*dev
)
132 struct net_bridge_port
*port
= rcu_dereference(dev
->br_port
);
134 return port
? port
->br
->dev
: NULL
;
137 static inline struct nf_bridge_info
*nf_bridge_alloc(struct sk_buff
*skb
)
139 skb
->nf_bridge
= kzalloc(sizeof(struct nf_bridge_info
), GFP_ATOMIC
);
140 if (likely(skb
->nf_bridge
))
141 atomic_set(&(skb
->nf_bridge
->use
), 1);
143 return skb
->nf_bridge
;
146 static inline struct nf_bridge_info
*nf_bridge_unshare(struct sk_buff
*skb
)
148 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
150 if (atomic_read(&nf_bridge
->use
) > 1) {
151 struct nf_bridge_info
*tmp
= nf_bridge_alloc(skb
);
154 memcpy(tmp
, nf_bridge
, sizeof(struct nf_bridge_info
));
155 atomic_set(&tmp
->use
, 1);
156 nf_bridge_put(nf_bridge
);
163 static inline void nf_bridge_push_encap_header(struct sk_buff
*skb
)
165 unsigned int len
= nf_bridge_encap_header_len(skb
);
168 skb
->network_header
-= len
;
171 static inline void nf_bridge_pull_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_rcsum(struct sk_buff
*skb
)
181 unsigned int len
= nf_bridge_encap_header_len(skb
);
183 skb_pull_rcsum(skb
, len
);
184 skb
->network_header
+= len
;
187 static inline void nf_bridge_save_header(struct sk_buff
*skb
)
189 int header_size
= ETH_HLEN
+ nf_bridge_encap_header_len(skb
);
191 skb_copy_from_linear_data_offset(skb
, -header_size
,
192 skb
->nf_bridge
->data
, header_size
);
196 * When forwarding bridge frames, we save a copy of the original
197 * header before processing.
199 int nf_bridge_copy_header(struct sk_buff
*skb
)
202 int header_size
= ETH_HLEN
+ nf_bridge_encap_header_len(skb
);
204 err
= skb_cow_head(skb
, header_size
);
208 skb_copy_to_linear_data_offset(skb
, -header_size
,
209 skb
->nf_bridge
->data
, header_size
);
210 __skb_push(skb
, nf_bridge_encap_header_len(skb
));
214 /* PF_BRIDGE/PRE_ROUTING *********************************************/
215 /* Undo the changes made for ip6tables PREROUTING and continue the
216 * bridge PRE_ROUTING hook. */
217 static int br_nf_pre_routing_finish_ipv6(struct sk_buff
*skb
)
219 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
221 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
222 skb
->pkt_type
= PACKET_OTHERHOST
;
223 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
225 nf_bridge
->mask
^= BRNF_NF_BRIDGE_PREROUTING
;
227 skb
->dst
= (struct dst_entry
*)&__fake_rtable
;
230 skb
->dev
= nf_bridge
->physindev
;
231 nf_bridge_push_encap_header(skb
);
232 NF_HOOK_THRESH(PF_BRIDGE
, NF_BR_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
233 br_handle_frame_finish
, 1);
238 static void __br_dnat_complain(void)
240 static unsigned long last_complaint
;
242 if (jiffies
- last_complaint
>= 5 * HZ
) {
243 printk(KERN_WARNING
"Performing cross-bridge DNAT requires IP "
244 "forwarding to be enabled\n");
245 last_complaint
= jiffies
;
249 /* This requires some explaining. If DNAT has taken place,
250 * we will need to fix up the destination Ethernet address,
251 * and this is a tricky process.
253 * There are two cases to consider:
254 * 1. The packet was DNAT'ed to a device in the same bridge
255 * port group as it was received on. We can still bridge
257 * 2. The packet was DNAT'ed to a different device, either
258 * a non-bridged device or another bridge port group.
259 * The packet will need to be routed.
261 * The correct way of distinguishing between these two cases is to
262 * call ip_route_input() and to look at skb->dst->dev, which is
263 * changed to the destination device if ip_route_input() succeeds.
265 * Let us first consider the case that ip_route_input() succeeds:
267 * If skb->dst->dev equals the logical bridge device the packet
268 * came in on, we can consider this bridging. The packet is passed
269 * through the neighbour output function to build a new destination
270 * MAC address, which will make the packet enter br_nf_local_out()
271 * not much later. In that function it is assured that the iptables
272 * FORWARD chain is traversed for the packet.
274 * Otherwise, the packet is considered to be routed and we just
275 * change the destination MAC address so that the packet will
276 * later be passed up to the IP stack to be routed. For a redirected
277 * packet, ip_route_input() will give back the localhost as output device,
278 * which differs from the bridge device.
280 * Let us now consider the case that ip_route_input() fails:
282 * This can be because the destination address is martian, in which case
283 * the packet will be dropped.
284 * After a "echo '0' > /proc/sys/net/ipv4/ip_forward" ip_route_input()
285 * will fail, while __ip_route_output_key() will return success. The source
286 * address for __ip_route_output_key() is set to zero, so __ip_route_output_key
287 * thinks we're handling a locally generated packet and won't care
288 * if IP forwarding is allowed. We send a warning message to the users's
289 * log telling her to put IP forwarding on.
291 * ip_route_input() will also fail if there is no route available.
292 * In that case we just drop the packet.
294 * --Lennert, 20020411
295 * --Bart, 20020416 (updated)
296 * --Bart, 20021007 (updated)
297 * --Bart, 20062711 (updated) */
298 static int br_nf_pre_routing_finish_bridge(struct sk_buff
*skb
)
300 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
301 skb
->pkt_type
= PACKET_HOST
;
302 skb
->nf_bridge
->mask
|= BRNF_PKT_TYPE
;
304 skb
->nf_bridge
->mask
^= BRNF_NF_BRIDGE_PREROUTING
;
306 skb
->dev
= bridge_parent(skb
->dev
);
308 struct dst_entry
*dst
= skb
->dst
;
310 nf_bridge_pull_encap_header(skb
);
313 return neigh_hh_output(dst
->hh
, skb
);
314 else if (dst
->neighbour
)
315 return dst
->neighbour
->output(skb
);
321 static int br_nf_pre_routing_finish(struct sk_buff
*skb
)
323 struct net_device
*dev
= skb
->dev
;
324 struct iphdr
*iph
= ip_hdr(skb
);
325 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
))) {
341 .tos
= RT_TOS(iph
->tos
) },
345 struct in_device
*in_dev
= in_dev_get(dev
);
347 /* If err equals -EHOSTUNREACH the error is due to a
348 * martian destination or due to the fact that
349 * forwarding is disabled. For most martian packets,
350 * ip_route_output_key() will fail. It won't fail for 2 types of
351 * martian destinations: loopback destinations and destination
352 * 0.0.0.0. In both cases the packet will be dropped because the
353 * destination is the loopback device and not the bridge. */
354 if (err
!= -EHOSTUNREACH
|| !in_dev
|| IN_DEV_FORWARD(in_dev
))
357 if (!ip_route_output_key(&init_net
, &rt
, &fl
)) {
358 /* - Bridged-and-DNAT'ed traffic doesn't
359 * require ip_forwarding. */
360 if (((struct dst_entry
*)rt
)->dev
== dev
) {
361 skb
->dst
= (struct dst_entry
*)rt
;
364 /* we are sure that forwarding is disabled, so printing
365 * this message is no problem. Note that the packet could
366 * still have a martian destination address, in which case
367 * the packet could be dropped even if forwarding were enabled */
368 __br_dnat_complain();
369 dst_release((struct dst_entry
*)rt
);
375 if (skb
->dst
->dev
== dev
) {
377 /* Tell br_nf_local_out this is a
379 nf_bridge
->mask
|= BRNF_BRIDGED_DNAT
;
380 skb
->dev
= nf_bridge
->physindev
;
381 nf_bridge_push_encap_header(skb
);
382 NF_HOOK_THRESH(PF_BRIDGE
, NF_BR_PRE_ROUTING
,
384 br_nf_pre_routing_finish_bridge
,
388 memcpy(eth_hdr(skb
)->h_dest
, dev
->dev_addr
, ETH_ALEN
);
389 skb
->pkt_type
= PACKET_HOST
;
392 skb
->dst
= (struct dst_entry
*)&__fake_rtable
;
396 skb
->dev
= nf_bridge
->physindev
;
397 nf_bridge_push_encap_header(skb
);
398 NF_HOOK_THRESH(PF_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(PF_INET6
, 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(PF_INET
, 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 if (skb
->dst
== (struct dst_entry
*)&__fake_rtable
) {
613 dst_release(skb
->dst
);
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(PF_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
) ||
673 nf_bridge_pull_encap_header(skb
);
675 nf_bridge
= skb
->nf_bridge
;
676 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
677 skb
->pkt_type
= PACKET_HOST
;
678 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
681 /* The physdev module checks on this */
682 nf_bridge
->mask
|= BRNF_BRIDGED
;
683 nf_bridge
->physoutdev
= skb
->dev
;
685 NF_HOOK(pf
, NF_INET_FORWARD
, skb
, bridge_parent(in
), parent
,
686 br_nf_forward_finish
);
691 static unsigned int br_nf_forward_arp(unsigned int hook
, struct sk_buff
*skb
,
692 const struct net_device
*in
,
693 const struct net_device
*out
,
694 int (*okfn
)(struct sk_buff
*))
696 struct net_device
**d
= (struct net_device
**)(skb
->cb
);
699 if (!brnf_call_arptables
)
703 if (skb
->protocol
!= htons(ETH_P_ARP
)) {
704 if (!IS_VLAN_ARP(skb
))
706 nf_bridge_pull_encap_header(skb
);
709 if (arp_hdr(skb
)->ar_pln
!= 4) {
710 if (IS_VLAN_ARP(skb
))
711 nf_bridge_push_encap_header(skb
);
714 *d
= (struct net_device
*)in
;
715 NF_HOOK(NF_ARP
, NF_ARP_FORWARD
, skb
, (struct net_device
*)in
,
716 (struct net_device
*)out
, br_nf_forward_finish
);
721 /* PF_BRIDGE/LOCAL_OUT ***********************************************
723 * This function sees both locally originated IP packets and forwarded
724 * IP packets (in both cases the destination device is a bridge
725 * device). It also sees bridged-and-DNAT'ed packets.
727 * If (nf_bridge->mask & BRNF_BRIDGED_DNAT) then the packet is bridged
728 * and we fake the PF_BRIDGE/FORWARD hook. The function br_nf_forward()
729 * will then fake the PF_INET/FORWARD hook. br_nf_local_out() has priority
730 * NF_BR_PRI_FIRST, so no relevant PF_BRIDGE/INPUT functions have been nor
733 static unsigned int br_nf_local_out(unsigned int hook
, struct sk_buff
*skb
,
734 const struct net_device
*in
,
735 const struct net_device
*out
,
736 int (*okfn
)(struct sk_buff
*))
738 struct net_device
*realindev
;
739 struct nf_bridge_info
*nf_bridge
;
744 /* Need exclusive nf_bridge_info since we might have multiple
745 * different physoutdevs. */
746 if (!nf_bridge_unshare(skb
))
749 nf_bridge
= skb
->nf_bridge
;
750 if (!(nf_bridge
->mask
& BRNF_BRIDGED_DNAT
))
753 /* Bridged, take PF_BRIDGE/FORWARD.
754 * (see big note in front of br_nf_pre_routing_finish) */
755 nf_bridge
->physoutdev
= skb
->dev
;
756 realindev
= nf_bridge
->physindev
;
758 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
759 skb
->pkt_type
= PACKET_OTHERHOST
;
760 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
762 nf_bridge_push_encap_header(skb
);
764 NF_HOOK(PF_BRIDGE
, NF_BR_FORWARD
, skb
, realindev
, skb
->dev
,
769 static int br_nf_dev_queue_xmit(struct sk_buff
*skb
)
771 if (skb
->protocol
== htons(ETH_P_IP
) &&
772 skb
->len
> skb
->dev
->mtu
&&
774 return ip_fragment(skb
, br_dev_queue_push_xmit
);
776 return br_dev_queue_push_xmit(skb
);
779 /* PF_BRIDGE/POST_ROUTING ********************************************/
780 static unsigned int br_nf_post_routing(unsigned int hook
, struct sk_buff
*skb
,
781 const struct net_device
*in
,
782 const struct net_device
*out
,
783 int (*okfn
)(struct sk_buff
*))
785 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
786 struct net_device
*realoutdev
= bridge_parent(skb
->dev
);
789 #ifdef CONFIG_NETFILTER_DEBUG
790 /* Be very paranoid. This probably won't happen anymore, but let's
791 * keep the check just to be sure... */
792 if (skb_mac_header(skb
) < skb
->head
||
793 skb_mac_header(skb
) + ETH_HLEN
> skb
->data
) {
794 printk(KERN_CRIT
"br_netfilter: Argh!! br_nf_post_routing: "
795 "bad mac.raw pointer.\n");
803 if (!(nf_bridge
->mask
& (BRNF_BRIDGED
| BRNF_BRIDGED_DNAT
)))
809 if (skb
->protocol
== htons(ETH_P_IP
) || IS_VLAN_IP(skb
) ||
815 #ifdef CONFIG_NETFILTER_DEBUG
816 if (skb
->dst
== NULL
) {
817 printk(KERN_INFO
"br_netfilter post_routing: skb->dst == NULL\n");
822 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
823 * about the value of skb->pkt_type. */
824 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
825 skb
->pkt_type
= PACKET_HOST
;
826 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
829 nf_bridge_pull_encap_header(skb
);
830 nf_bridge_save_header(skb
);
832 NF_HOOK(pf
, NF_INET_POST_ROUTING
, skb
, NULL
, realoutdev
,
833 br_nf_dev_queue_xmit
);
837 #ifdef CONFIG_NETFILTER_DEBUG
839 if (skb
->dev
!= NULL
) {
840 printk("[%s]", skb
->dev
->name
);
842 printk("[%s]", realoutdev
->name
);
844 printk(" head:%p, raw:%p, data:%p\n", skb
->head
, skb_mac_header(skb
),
851 /* IP/SABOTAGE *****************************************************/
852 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
853 * for the second time. */
854 static unsigned int ip_sabotage_in(unsigned int hook
, struct sk_buff
*skb
,
855 const struct net_device
*in
,
856 const struct net_device
*out
,
857 int (*okfn
)(struct sk_buff
*))
859 if (skb
->nf_bridge
&&
860 !(skb
->nf_bridge
->mask
& BRNF_NF_BRIDGE_PREROUTING
)) {
867 /* For br_nf_local_out we need (prio = NF_BR_PRI_FIRST), to insure that innocent
868 * PF_BRIDGE/NF_BR_LOCAL_OUT functions don't get bridged traffic as input.
869 * For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
870 * ip_refrag() can return NF_STOLEN. */
871 static struct nf_hook_ops br_nf_ops
[] __read_mostly
= {
872 { .hook
= br_nf_pre_routing
,
873 .owner
= THIS_MODULE
,
875 .hooknum
= NF_BR_PRE_ROUTING
,
876 .priority
= NF_BR_PRI_BRNF
, },
877 { .hook
= br_nf_local_in
,
878 .owner
= THIS_MODULE
,
880 .hooknum
= NF_BR_LOCAL_IN
,
881 .priority
= NF_BR_PRI_BRNF
, },
882 { .hook
= br_nf_forward_ip
,
883 .owner
= THIS_MODULE
,
885 .hooknum
= NF_BR_FORWARD
,
886 .priority
= NF_BR_PRI_BRNF
- 1, },
887 { .hook
= br_nf_forward_arp
,
888 .owner
= THIS_MODULE
,
890 .hooknum
= NF_BR_FORWARD
,
891 .priority
= NF_BR_PRI_BRNF
, },
892 { .hook
= br_nf_local_out
,
893 .owner
= THIS_MODULE
,
895 .hooknum
= NF_BR_LOCAL_OUT
,
896 .priority
= NF_BR_PRI_FIRST
, },
897 { .hook
= br_nf_post_routing
,
898 .owner
= THIS_MODULE
,
900 .hooknum
= NF_BR_POST_ROUTING
,
901 .priority
= NF_BR_PRI_LAST
, },
902 { .hook
= ip_sabotage_in
,
903 .owner
= THIS_MODULE
,
905 .hooknum
= NF_INET_PRE_ROUTING
,
906 .priority
= NF_IP_PRI_FIRST
, },
907 { .hook
= ip_sabotage_in
,
908 .owner
= THIS_MODULE
,
910 .hooknum
= NF_INET_PRE_ROUTING
,
911 .priority
= NF_IP6_PRI_FIRST
, },
916 int brnf_sysctl_call_tables(ctl_table
* ctl
, int write
, struct file
*filp
,
917 void __user
* buffer
, size_t * lenp
, loff_t
* ppos
)
921 ret
= proc_dointvec(ctl
, write
, filp
, buffer
, lenp
, ppos
);
923 if (write
&& *(int *)(ctl
->data
))
924 *(int *)(ctl
->data
) = 1;
928 static ctl_table brnf_table
[] = {
930 .procname
= "bridge-nf-call-arptables",
931 .data
= &brnf_call_arptables
,
932 .maxlen
= sizeof(int),
934 .proc_handler
= &brnf_sysctl_call_tables
,
937 .procname
= "bridge-nf-call-iptables",
938 .data
= &brnf_call_iptables
,
939 .maxlen
= sizeof(int),
941 .proc_handler
= &brnf_sysctl_call_tables
,
944 .procname
= "bridge-nf-call-ip6tables",
945 .data
= &brnf_call_ip6tables
,
946 .maxlen
= sizeof(int),
948 .proc_handler
= &brnf_sysctl_call_tables
,
951 .procname
= "bridge-nf-filter-vlan-tagged",
952 .data
= &brnf_filter_vlan_tagged
,
953 .maxlen
= sizeof(int),
955 .proc_handler
= &brnf_sysctl_call_tables
,
958 .procname
= "bridge-nf-filter-pppoe-tagged",
959 .data
= &brnf_filter_pppoe_tagged
,
960 .maxlen
= sizeof(int),
962 .proc_handler
= &brnf_sysctl_call_tables
,
967 static struct ctl_path brnf_path
[] = {
968 { .procname
= "net", .ctl_name
= CTL_NET
, },
969 { .procname
= "bridge", .ctl_name
= NET_BRIDGE
, },
974 int __init
br_netfilter_init(void)
978 ret
= nf_register_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
982 brnf_sysctl_header
= register_sysctl_paths(brnf_path
, brnf_table
);
983 if (brnf_sysctl_header
== NULL
) {
985 "br_netfilter: can't register to sysctl.\n");
986 nf_unregister_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
990 printk(KERN_NOTICE
"Bridge firewalling registered\n");
994 void br_netfilter_fini(void)
996 nf_unregister_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
998 unregister_sysctl_table(brnf_sysctl_header
);