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 if (!br_port_exists(dev
))
135 return &br_port_get_rcu(dev
)->br
->fake_rtable
;
138 static inline struct net_device
*bridge_parent(const struct net_device
*dev
)
140 if (!br_port_exists(dev
))
143 return br_port_get_rcu(dev
)->br
->dev
;
146 static inline struct nf_bridge_info
*nf_bridge_alloc(struct sk_buff
*skb
)
148 skb
->nf_bridge
= kzalloc(sizeof(struct nf_bridge_info
), GFP_ATOMIC
);
149 if (likely(skb
->nf_bridge
))
150 atomic_set(&(skb
->nf_bridge
->use
), 1);
152 return skb
->nf_bridge
;
155 static inline struct nf_bridge_info
*nf_bridge_unshare(struct sk_buff
*skb
)
157 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
159 if (atomic_read(&nf_bridge
->use
) > 1) {
160 struct nf_bridge_info
*tmp
= nf_bridge_alloc(skb
);
163 memcpy(tmp
, nf_bridge
, sizeof(struct nf_bridge_info
));
164 atomic_set(&tmp
->use
, 1);
166 nf_bridge_put(nf_bridge
);
172 static inline void nf_bridge_push_encap_header(struct sk_buff
*skb
)
174 unsigned int len
= nf_bridge_encap_header_len(skb
);
177 skb
->network_header
-= len
;
180 static inline void nf_bridge_pull_encap_header(struct sk_buff
*skb
)
182 unsigned int len
= nf_bridge_encap_header_len(skb
);
185 skb
->network_header
+= len
;
188 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff
*skb
)
190 unsigned int len
= nf_bridge_encap_header_len(skb
);
192 skb_pull_rcsum(skb
, len
);
193 skb
->network_header
+= len
;
196 static inline void nf_bridge_save_header(struct sk_buff
*skb
)
198 int header_size
= ETH_HLEN
+ nf_bridge_encap_header_len(skb
);
200 skb_copy_from_linear_data_offset(skb
, -header_size
,
201 skb
->nf_bridge
->data
, header_size
);
204 static inline void nf_bridge_update_protocol(struct sk_buff
*skb
)
206 if (skb
->nf_bridge
->mask
& BRNF_8021Q
)
207 skb
->protocol
= htons(ETH_P_8021Q
);
208 else if (skb
->nf_bridge
->mask
& BRNF_PPPoE
)
209 skb
->protocol
= htons(ETH_P_PPP_SES
);
212 /* Fill in the header for fragmented IP packets handled by
213 * the IPv4 connection tracking code.
215 int nf_bridge_copy_header(struct sk_buff
*skb
)
218 unsigned int header_size
;
220 nf_bridge_update_protocol(skb
);
221 header_size
= ETH_HLEN
+ nf_bridge_encap_header_len(skb
);
222 err
= skb_cow_head(skb
, header_size
);
226 skb_copy_to_linear_data_offset(skb
, -header_size
,
227 skb
->nf_bridge
->data
, header_size
);
228 __skb_push(skb
, nf_bridge_encap_header_len(skb
));
232 /* PF_BRIDGE/PRE_ROUTING *********************************************/
233 /* Undo the changes made for ip6tables PREROUTING and continue the
234 * bridge PRE_ROUTING hook. */
235 static int br_nf_pre_routing_finish_ipv6(struct sk_buff
*skb
)
237 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
240 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
241 skb
->pkt_type
= PACKET_OTHERHOST
;
242 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
244 nf_bridge
->mask
^= BRNF_NF_BRIDGE_PREROUTING
;
246 rt
= bridge_parent_rtable(nf_bridge
->physindev
);
251 skb_dst_set_noref(skb
, &rt
->dst
);
253 skb
->dev
= nf_bridge
->physindev
;
254 nf_bridge_update_protocol(skb
);
255 nf_bridge_push_encap_header(skb
);
256 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
257 br_handle_frame_finish
, 1);
262 /* Obtain the correct destination MAC address, while preserving the original
263 * source MAC address. If we already know this address, we just copy it. If we
264 * don't, we use the neighbour framework to find out. In both cases, we make
265 * sure that br_handle_frame_finish() is called afterwards.
267 static int br_nf_pre_routing_finish_bridge(struct sk_buff
*skb
)
269 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
270 struct dst_entry
*dst
;
272 skb
->dev
= bridge_parent(skb
->dev
);
277 neigh_hh_bridge(dst
->hh
, skb
);
278 skb
->dev
= nf_bridge
->physindev
;
279 return br_handle_frame_finish(skb
);
280 } else if (dst
->neighbour
) {
281 /* the neighbour function below overwrites the complete
282 * MAC header, so we save the Ethernet source address and
283 * protocol number. */
284 skb_copy_from_linear_data_offset(skb
, -(ETH_HLEN
-ETH_ALEN
), skb
->nf_bridge
->data
, ETH_HLEN
-ETH_ALEN
);
285 /* tell br_dev_xmit to continue with forwarding */
286 nf_bridge
->mask
|= BRNF_BRIDGED_DNAT
;
287 return dst
->neighbour
->output(skb
);
294 /* This requires some explaining. If DNAT has taken place,
295 * we will need to fix up the destination Ethernet address.
297 * There are two cases to consider:
298 * 1. The packet was DNAT'ed to a device in the same bridge
299 * port group as it was received on. We can still bridge
301 * 2. The packet was DNAT'ed to a different device, either
302 * a non-bridged device or another bridge port group.
303 * The packet will need to be routed.
305 * The correct way of distinguishing between these two cases is to
306 * call ip_route_input() and to look at skb->dst->dev, which is
307 * changed to the destination device if ip_route_input() succeeds.
309 * Let's first consider the case that ip_route_input() succeeds:
311 * If the output device equals the logical bridge device the packet
312 * came in on, we can consider this bridging. The corresponding MAC
313 * address will be obtained in br_nf_pre_routing_finish_bridge.
314 * Otherwise, the packet is considered to be routed and we just
315 * change the destination MAC address so that the packet will
316 * later be passed up to the IP stack to be routed. For a redirected
317 * packet, ip_route_input() will give back the localhost as output device,
318 * which differs from the bridge device.
320 * Let's now consider the case that ip_route_input() fails:
322 * This can be because the destination address is martian, in which case
323 * the packet will be dropped.
324 * If IP forwarding is disabled, ip_route_input() will fail, while
325 * ip_route_output_key() can return success. The source
326 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
327 * thinks we're handling a locally generated packet and won't care
328 * if IP forwarding is enabled. If the output device equals the logical bridge
329 * device, we proceed as if ip_route_input() succeeded. If it differs from the
330 * logical bridge port or if ip_route_output_key() fails we drop the packet.
332 static int br_nf_pre_routing_finish(struct sk_buff
*skb
)
334 struct net_device
*dev
= skb
->dev
;
335 struct iphdr
*iph
= ip_hdr(skb
);
336 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
340 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
341 skb
->pkt_type
= PACKET_OTHERHOST
;
342 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
344 nf_bridge
->mask
^= BRNF_NF_BRIDGE_PREROUTING
;
345 if (dnat_took_place(skb
)) {
346 if ((err
= ip_route_input(skb
, iph
->daddr
, iph
->saddr
, iph
->tos
, dev
))) {
352 .tos
= RT_TOS(iph
->tos
) },
356 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
358 /* If err equals -EHOSTUNREACH the error is due to a
359 * martian destination or due to the fact that
360 * forwarding is disabled. For most martian packets,
361 * ip_route_output_key() will fail. It won't fail for 2 types of
362 * martian destinations: loopback destinations and destination
363 * 0.0.0.0. In both cases the packet will be dropped because the
364 * destination is the loopback device and not the bridge. */
365 if (err
!= -EHOSTUNREACH
|| !in_dev
|| IN_DEV_FORWARD(in_dev
))
368 if (!ip_route_output_key(dev_net(dev
), &rt
, &fl
)) {
369 /* - Bridged-and-DNAT'ed traffic doesn't
370 * require ip_forwarding. */
371 if (((struct dst_entry
*)rt
)->dev
== dev
) {
372 skb_dst_set(skb
, (struct dst_entry
*)rt
);
375 dst_release((struct dst_entry
*)rt
);
381 if (skb_dst(skb
)->dev
== dev
) {
383 skb
->dev
= nf_bridge
->physindev
;
384 nf_bridge_update_protocol(skb
);
385 nf_bridge_push_encap_header(skb
);
386 NF_HOOK_THRESH(NFPROTO_BRIDGE
,
389 br_nf_pre_routing_finish_bridge
,
393 memcpy(eth_hdr(skb
)->h_dest
, dev
->dev_addr
, ETH_ALEN
);
394 skb
->pkt_type
= PACKET_HOST
;
397 rt
= bridge_parent_rtable(nf_bridge
->physindev
);
402 skb_dst_set_noref(skb
, &rt
->dst
);
405 skb
->dev
= nf_bridge
->physindev
;
406 nf_bridge_update_protocol(skb
);
407 nf_bridge_push_encap_header(skb
);
408 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
409 br_handle_frame_finish
, 1);
414 /* Some common code for IPv4/IPv6 */
415 static struct net_device
*setup_pre_routing(struct sk_buff
*skb
)
417 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
419 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
420 skb
->pkt_type
= PACKET_HOST
;
421 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
424 nf_bridge
->mask
|= BRNF_NF_BRIDGE_PREROUTING
;
425 nf_bridge
->physindev
= skb
->dev
;
426 skb
->dev
= bridge_parent(skb
->dev
);
427 if (skb
->protocol
== htons(ETH_P_8021Q
))
428 nf_bridge
->mask
|= BRNF_8021Q
;
429 else if (skb
->protocol
== htons(ETH_P_PPP_SES
))
430 nf_bridge
->mask
|= BRNF_PPPoE
;
435 /* We only check the length. A bridge shouldn't do any hop-by-hop stuff anyway */
436 static int check_hbh_len(struct sk_buff
*skb
)
438 unsigned char *raw
= (u8
*)(ipv6_hdr(skb
) + 1);
440 const unsigned char *nh
= skb_network_header(skb
);
442 int len
= (raw
[1] + 1) << 3;
444 if ((raw
+ len
) - skb
->data
> skb_headlen(skb
))
451 int optlen
= nh
[off
+ 1] + 2;
462 if (nh
[off
+ 1] != 4 || (off
& 3) != 2)
464 pkt_len
= ntohl(*(__be32
*) (nh
+ off
+ 2));
465 if (pkt_len
<= IPV6_MAXPLEN
||
466 ipv6_hdr(skb
)->payload_len
)
468 if (pkt_len
> skb
->len
- sizeof(struct ipv6hdr
))
470 if (pskb_trim_rcsum(skb
,
471 pkt_len
+ sizeof(struct ipv6hdr
)))
473 nh
= skb_network_header(skb
);
490 /* Replicate the checks that IPv6 does on packet reception and pass the packet
491 * to ip6tables, which doesn't support NAT, so things are fairly simple. */
492 static unsigned int br_nf_pre_routing_ipv6(unsigned int hook
,
494 const struct net_device
*in
,
495 const struct net_device
*out
,
496 int (*okfn
)(struct sk_buff
*))
501 if (skb
->len
< sizeof(struct ipv6hdr
))
504 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
509 if (hdr
->version
!= 6)
512 pkt_len
= ntohs(hdr
->payload_len
);
514 if (pkt_len
|| hdr
->nexthdr
!= NEXTHDR_HOP
) {
515 if (pkt_len
+ sizeof(struct ipv6hdr
) > skb
->len
)
517 if (pskb_trim_rcsum(skb
, pkt_len
+ sizeof(struct ipv6hdr
)))
520 if (hdr
->nexthdr
== NEXTHDR_HOP
&& check_hbh_len(skb
))
523 nf_bridge_put(skb
->nf_bridge
);
524 if (!nf_bridge_alloc(skb
))
526 if (!setup_pre_routing(skb
))
529 skb
->protocol
= htons(ETH_P_IPV6
);
530 NF_HOOK(NFPROTO_IPV6
, NF_INET_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
531 br_nf_pre_routing_finish_ipv6
);
539 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
540 * Replicate the checks that IPv4 does on packet reception.
541 * Set skb->dev to the bridge device (i.e. parent of the
542 * receiving device) to make netfilter happy, the REDIRECT
543 * target in particular. Save the original destination IP
544 * address to be able to detect DNAT afterwards. */
545 static unsigned int br_nf_pre_routing(unsigned int hook
, struct sk_buff
*skb
,
546 const struct net_device
*in
,
547 const struct net_device
*out
,
548 int (*okfn
)(struct sk_buff
*))
550 struct net_bridge_port
*p
;
551 struct net_bridge
*br
;
553 __u32 len
= nf_bridge_encap_header_len(skb
);
555 if (unlikely(!pskb_may_pull(skb
, len
)))
558 p
= br_port_get_rcu(in
);
563 if (skb
->protocol
== htons(ETH_P_IPV6
) || IS_VLAN_IPV6(skb
) ||
564 IS_PPPOE_IPV6(skb
)) {
565 if (!brnf_call_ip6tables
&& !br
->nf_call_ip6tables
)
568 nf_bridge_pull_encap_header_rcsum(skb
);
569 return br_nf_pre_routing_ipv6(hook
, skb
, in
, out
, okfn
);
572 if (!brnf_call_iptables
&& !br
->nf_call_iptables
)
575 if (skb
->protocol
!= htons(ETH_P_IP
) && !IS_VLAN_IP(skb
) &&
579 nf_bridge_pull_encap_header_rcsum(skb
);
581 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
585 if (iph
->ihl
< 5 || iph
->version
!= 4)
588 if (!pskb_may_pull(skb
, 4 * iph
->ihl
))
592 if (ip_fast_csum((__u8
*) iph
, iph
->ihl
) != 0)
595 len
= ntohs(iph
->tot_len
);
596 if (skb
->len
< len
|| len
< 4 * iph
->ihl
)
599 pskb_trim_rcsum(skb
, len
);
601 /* BUG: Should really parse the IP options here. */
602 memset(IPCB(skb
), 0, sizeof(struct inet_skb_parm
));
604 nf_bridge_put(skb
->nf_bridge
);
605 if (!nf_bridge_alloc(skb
))
607 if (!setup_pre_routing(skb
))
609 store_orig_dstaddr(skb
);
610 skb
->protocol
= htons(ETH_P_IP
);
612 NF_HOOK(NFPROTO_IPV4
, NF_INET_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
613 br_nf_pre_routing_finish
);
618 // IP_INC_STATS_BH(IpInHdrErrors);
624 /* PF_BRIDGE/LOCAL_IN ************************************************/
625 /* The packet is locally destined, which requires a real
626 * dst_entry, so detach the fake one. On the way up, the
627 * packet would pass through PRE_ROUTING again (which already
628 * took place when the packet entered the bridge), but we
629 * register an IPv4 PRE_ROUTING 'sabotage' hook that will
630 * prevent this from happening. */
631 static unsigned int br_nf_local_in(unsigned int hook
, struct sk_buff
*skb
,
632 const struct net_device
*in
,
633 const struct net_device
*out
,
634 int (*okfn
)(struct sk_buff
*))
636 struct rtable
*rt
= skb_rtable(skb
);
638 if (rt
&& rt
== bridge_parent_rtable(in
))
644 /* PF_BRIDGE/FORWARD *************************************************/
645 static int br_nf_forward_finish(struct sk_buff
*skb
)
647 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
648 struct net_device
*in
;
650 if (skb
->protocol
!= htons(ETH_P_ARP
) && !IS_VLAN_ARP(skb
)) {
651 in
= nf_bridge
->physindev
;
652 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
653 skb
->pkt_type
= PACKET_OTHERHOST
;
654 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
656 nf_bridge_update_protocol(skb
);
658 in
= *((struct net_device
**)(skb
->cb
));
660 nf_bridge_push_encap_header(skb
);
662 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_FORWARD
, skb
, in
,
663 skb
->dev
, br_forward_finish
, 1);
667 /* This is the 'purely bridged' case. For IP, we pass the packet to
668 * netfilter with indev and outdev set to the bridge device,
669 * but we are still able to filter on the 'real' indev/outdev
670 * because of the physdev module. For ARP, indev and outdev are the
672 static unsigned int br_nf_forward_ip(unsigned int hook
, struct sk_buff
*skb
,
673 const struct net_device
*in
,
674 const struct net_device
*out
,
675 int (*okfn
)(struct sk_buff
*))
677 struct nf_bridge_info
*nf_bridge
;
678 struct net_device
*parent
;
684 /* Need exclusive nf_bridge_info since we might have multiple
685 * different physoutdevs. */
686 if (!nf_bridge_unshare(skb
))
689 parent
= bridge_parent(out
);
693 if (skb
->protocol
== htons(ETH_P_IP
) || IS_VLAN_IP(skb
) ||
696 else if (skb
->protocol
== htons(ETH_P_IPV6
) || IS_VLAN_IPV6(skb
) ||
702 nf_bridge_pull_encap_header(skb
);
704 nf_bridge
= skb
->nf_bridge
;
705 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
706 skb
->pkt_type
= PACKET_HOST
;
707 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
710 /* The physdev module checks on this */
711 nf_bridge
->mask
|= BRNF_BRIDGED
;
712 nf_bridge
->physoutdev
= skb
->dev
;
714 skb
->protocol
= htons(ETH_P_IP
);
716 skb
->protocol
= htons(ETH_P_IPV6
);
718 NF_HOOK(pf
, NF_INET_FORWARD
, skb
, bridge_parent(in
), parent
,
719 br_nf_forward_finish
);
724 static unsigned int br_nf_forward_arp(unsigned int hook
, struct sk_buff
*skb
,
725 const struct net_device
*in
,
726 const struct net_device
*out
,
727 int (*okfn
)(struct sk_buff
*))
729 struct net_bridge_port
*p
;
730 struct net_bridge
*br
;
731 struct net_device
**d
= (struct net_device
**)(skb
->cb
);
733 p
= br_port_get_rcu(out
);
738 if (!brnf_call_arptables
&& !br
->nf_call_arptables
)
741 if (skb
->protocol
!= htons(ETH_P_ARP
)) {
742 if (!IS_VLAN_ARP(skb
))
744 nf_bridge_pull_encap_header(skb
);
747 if (arp_hdr(skb
)->ar_pln
!= 4) {
748 if (IS_VLAN_ARP(skb
))
749 nf_bridge_push_encap_header(skb
);
752 *d
= (struct net_device
*)in
;
753 NF_HOOK(NFPROTO_ARP
, NF_ARP_FORWARD
, skb
, (struct net_device
*)in
,
754 (struct net_device
*)out
, br_nf_forward_finish
);
759 #if defined(CONFIG_NF_CONNTRACK_IPV4) || defined(CONFIG_NF_CONNTRACK_IPV4_MODULE)
760 static int br_nf_dev_queue_xmit(struct sk_buff
*skb
)
762 if (skb
->nfct
!= NULL
&& skb
->protocol
== htons(ETH_P_IP
) &&
763 skb
->len
+ nf_bridge_mtu_reduction(skb
) > skb
->dev
->mtu
&&
765 /* BUG: Should really parse the IP options here. */
766 memset(IPCB(skb
), 0, sizeof(struct inet_skb_parm
));
767 return ip_fragment(skb
, br_dev_queue_push_xmit
);
769 return br_dev_queue_push_xmit(skb
);
772 static int br_nf_dev_queue_xmit(struct sk_buff
*skb
)
774 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 if (!nf_bridge
|| !(nf_bridge
->mask
& BRNF_BRIDGED
))
794 if (skb
->protocol
== htons(ETH_P_IP
) || IS_VLAN_IP(skb
) ||
797 else if (skb
->protocol
== htons(ETH_P_IPV6
) || IS_VLAN_IPV6(skb
) ||
803 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
804 * about the value of skb->pkt_type. */
805 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
806 skb
->pkt_type
= PACKET_HOST
;
807 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
810 nf_bridge_pull_encap_header(skb
);
811 nf_bridge_save_header(skb
);
813 skb
->protocol
= htons(ETH_P_IP
);
815 skb
->protocol
= htons(ETH_P_IPV6
);
817 NF_HOOK(pf
, NF_INET_POST_ROUTING
, skb
, NULL
, realoutdev
,
818 br_nf_dev_queue_xmit
);
823 /* IP/SABOTAGE *****************************************************/
824 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
825 * for the second time. */
826 static unsigned int ip_sabotage_in(unsigned int hook
, struct sk_buff
*skb
,
827 const struct net_device
*in
,
828 const struct net_device
*out
,
829 int (*okfn
)(struct sk_buff
*))
831 if (skb
->nf_bridge
&&
832 !(skb
->nf_bridge
->mask
& BRNF_NF_BRIDGE_PREROUTING
)) {
839 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
840 * br_dev_queue_push_xmit is called afterwards */
841 static struct nf_hook_ops br_nf_ops
[] __read_mostly
= {
843 .hook
= br_nf_pre_routing
,
844 .owner
= THIS_MODULE
,
846 .hooknum
= NF_BR_PRE_ROUTING
,
847 .priority
= NF_BR_PRI_BRNF
,
850 .hook
= br_nf_local_in
,
851 .owner
= THIS_MODULE
,
853 .hooknum
= NF_BR_LOCAL_IN
,
854 .priority
= NF_BR_PRI_BRNF
,
857 .hook
= br_nf_forward_ip
,
858 .owner
= THIS_MODULE
,
860 .hooknum
= NF_BR_FORWARD
,
861 .priority
= NF_BR_PRI_BRNF
- 1,
864 .hook
= br_nf_forward_arp
,
865 .owner
= THIS_MODULE
,
867 .hooknum
= NF_BR_FORWARD
,
868 .priority
= NF_BR_PRI_BRNF
,
871 .hook
= br_nf_post_routing
,
872 .owner
= THIS_MODULE
,
874 .hooknum
= NF_BR_POST_ROUTING
,
875 .priority
= NF_BR_PRI_LAST
,
878 .hook
= ip_sabotage_in
,
879 .owner
= THIS_MODULE
,
881 .hooknum
= NF_INET_PRE_ROUTING
,
882 .priority
= NF_IP_PRI_FIRST
,
885 .hook
= ip_sabotage_in
,
886 .owner
= THIS_MODULE
,
888 .hooknum
= NF_INET_PRE_ROUTING
,
889 .priority
= NF_IP6_PRI_FIRST
,
895 int brnf_sysctl_call_tables(ctl_table
* ctl
, int write
,
896 void __user
* buffer
, size_t * lenp
, loff_t
* ppos
)
900 ret
= proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
902 if (write
&& *(int *)(ctl
->data
))
903 *(int *)(ctl
->data
) = 1;
907 static ctl_table brnf_table
[] = {
909 .procname
= "bridge-nf-call-arptables",
910 .data
= &brnf_call_arptables
,
911 .maxlen
= sizeof(int),
913 .proc_handler
= brnf_sysctl_call_tables
,
916 .procname
= "bridge-nf-call-iptables",
917 .data
= &brnf_call_iptables
,
918 .maxlen
= sizeof(int),
920 .proc_handler
= brnf_sysctl_call_tables
,
923 .procname
= "bridge-nf-call-ip6tables",
924 .data
= &brnf_call_ip6tables
,
925 .maxlen
= sizeof(int),
927 .proc_handler
= brnf_sysctl_call_tables
,
930 .procname
= "bridge-nf-filter-vlan-tagged",
931 .data
= &brnf_filter_vlan_tagged
,
932 .maxlen
= sizeof(int),
934 .proc_handler
= brnf_sysctl_call_tables
,
937 .procname
= "bridge-nf-filter-pppoe-tagged",
938 .data
= &brnf_filter_pppoe_tagged
,
939 .maxlen
= sizeof(int),
941 .proc_handler
= brnf_sysctl_call_tables
,
946 static struct ctl_path brnf_path
[] = {
947 { .procname
= "net", },
948 { .procname
= "bridge", },
953 int __init
br_netfilter_init(void)
957 ret
= nf_register_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
961 brnf_sysctl_header
= register_sysctl_paths(brnf_path
, brnf_table
);
962 if (brnf_sysctl_header
== NULL
) {
964 "br_netfilter: can't register to sysctl.\n");
965 nf_unregister_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
969 printk(KERN_NOTICE
"Bridge firewalling registered\n");
973 void br_netfilter_fini(void)
975 nf_unregister_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
977 unregister_sysctl_table(brnf_sysctl_header
);