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>
38 #include <net/netfilter/br_netfilter.h>
40 #include <asm/uaccess.h>
41 #include "br_private.h"
43 #include <linux/sysctl.h>
46 #define skb_origaddr(skb) (((struct bridge_skb_cb *) \
47 (skb->nf_bridge->data))->daddr.ipv4)
48 #define store_orig_dstaddr(skb) (skb_origaddr(skb) = ip_hdr(skb)->daddr)
49 #define dnat_took_place(skb) (skb_origaddr(skb) != ip_hdr(skb)->daddr)
52 static struct ctl_table_header
*brnf_sysctl_header
;
53 static int brnf_call_iptables __read_mostly
= 1;
54 static int brnf_call_ip6tables __read_mostly
= 1;
55 static int brnf_call_arptables __read_mostly
= 1;
56 static int brnf_filter_vlan_tagged __read_mostly
= 0;
57 static int brnf_filter_pppoe_tagged __read_mostly
= 0;
58 static int brnf_pass_vlan_indev __read_mostly
= 0;
60 #define brnf_call_iptables 1
61 #define brnf_call_ip6tables 1
62 #define brnf_call_arptables 1
63 #define brnf_filter_vlan_tagged 0
64 #define brnf_filter_pppoe_tagged 0
65 #define brnf_pass_vlan_indev 0
69 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP))
71 #define IS_IPV6(skb) \
72 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6))
75 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP))
77 static inline __be16
vlan_proto(const struct sk_buff
*skb
)
79 if (skb_vlan_tag_present(skb
))
81 else if (skb
->protocol
== htons(ETH_P_8021Q
))
82 return vlan_eth_hdr(skb
)->h_vlan_encapsulated_proto
;
87 #define IS_VLAN_IP(skb) \
88 (vlan_proto(skb) == htons(ETH_P_IP) && \
89 brnf_filter_vlan_tagged)
91 #define IS_VLAN_IPV6(skb) \
92 (vlan_proto(skb) == htons(ETH_P_IPV6) && \
93 brnf_filter_vlan_tagged)
95 #define IS_VLAN_ARP(skb) \
96 (vlan_proto(skb) == htons(ETH_P_ARP) && \
97 brnf_filter_vlan_tagged)
99 static inline __be16
pppoe_proto(const struct sk_buff
*skb
)
101 return *((__be16
*)(skb_mac_header(skb
) + ETH_HLEN
+
102 sizeof(struct pppoe_hdr
)));
105 #define IS_PPPOE_IP(skb) \
106 (skb->protocol == htons(ETH_P_PPP_SES) && \
107 pppoe_proto(skb) == htons(PPP_IP) && \
108 brnf_filter_pppoe_tagged)
110 #define IS_PPPOE_IPV6(skb) \
111 (skb->protocol == htons(ETH_P_PPP_SES) && \
112 pppoe_proto(skb) == htons(PPP_IPV6) && \
113 brnf_filter_pppoe_tagged)
115 static inline struct rtable
*bridge_parent_rtable(const struct net_device
*dev
)
117 struct net_bridge_port
*port
;
119 port
= br_port_get_rcu(dev
);
120 return port
? &port
->br
->fake_rtable
: NULL
;
123 static inline struct net_device
*bridge_parent(const struct net_device
*dev
)
125 struct net_bridge_port
*port
;
127 port
= br_port_get_rcu(dev
);
128 return port
? port
->br
->dev
: NULL
;
131 static inline struct nf_bridge_info
*nf_bridge_alloc(struct sk_buff
*skb
)
133 skb
->nf_bridge
= kzalloc(sizeof(struct nf_bridge_info
), GFP_ATOMIC
);
134 if (likely(skb
->nf_bridge
))
135 atomic_set(&(skb
->nf_bridge
->use
), 1);
137 return skb
->nf_bridge
;
140 static inline struct nf_bridge_info
*nf_bridge_unshare(struct sk_buff
*skb
)
142 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
144 if (atomic_read(&nf_bridge
->use
) > 1) {
145 struct nf_bridge_info
*tmp
= nf_bridge_alloc(skb
);
148 memcpy(tmp
, nf_bridge
, sizeof(struct nf_bridge_info
));
149 atomic_set(&tmp
->use
, 1);
151 nf_bridge_put(nf_bridge
);
157 static inline void nf_bridge_push_encap_header(struct sk_buff
*skb
)
159 unsigned int len
= nf_bridge_encap_header_len(skb
);
162 skb
->network_header
-= len
;
165 static inline void nf_bridge_pull_encap_header(struct sk_buff
*skb
)
167 unsigned int len
= nf_bridge_encap_header_len(skb
);
170 skb
->network_header
+= len
;
173 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff
*skb
)
175 unsigned int len
= nf_bridge_encap_header_len(skb
);
177 skb_pull_rcsum(skb
, len
);
178 skb
->network_header
+= len
;
181 static inline void nf_bridge_save_header(struct sk_buff
*skb
)
183 int header_size
= ETH_HLEN
+ nf_bridge_encap_header_len(skb
);
185 skb_copy_from_linear_data_offset(skb
, -header_size
,
186 skb
->nf_bridge
->data
, header_size
);
189 /* When handing a packet over to the IP layer
190 * check whether we have a skb that is in the
194 static int br_parse_ip_options(struct sk_buff
*skb
)
196 const struct iphdr
*iph
;
197 struct net_device
*dev
= skb
->dev
;
200 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
205 /* Basic sanity checks */
206 if (iph
->ihl
< 5 || iph
->version
!= 4)
209 if (!pskb_may_pull(skb
, iph
->ihl
*4))
213 if (unlikely(ip_fast_csum((u8
*)iph
, iph
->ihl
)))
216 len
= ntohs(iph
->tot_len
);
217 if (skb
->len
< len
) {
218 IP_INC_STATS_BH(dev_net(dev
), IPSTATS_MIB_INTRUNCATEDPKTS
);
220 } else if (len
< (iph
->ihl
*4))
223 if (pskb_trim_rcsum(skb
, len
)) {
224 IP_INC_STATS_BH(dev_net(dev
), IPSTATS_MIB_INDISCARDS
);
228 memset(IPCB(skb
), 0, sizeof(struct inet_skb_parm
));
229 /* We should really parse IP options here but until
230 * somebody who actually uses IP options complains to
231 * us we'll just silently ignore the options because
237 IP_INC_STATS_BH(dev_net(dev
), IPSTATS_MIB_INHDRERRORS
);
242 /* PF_BRIDGE/PRE_ROUTING *********************************************/
243 /* Undo the changes made for ip6tables PREROUTING and continue the
244 * bridge PRE_ROUTING hook. */
245 static int br_nf_pre_routing_finish_ipv6(struct sk_buff
*skb
)
247 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
250 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
251 skb
->pkt_type
= PACKET_OTHERHOST
;
252 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
254 nf_bridge
->mask
^= BRNF_NF_BRIDGE_PREROUTING
;
256 rt
= bridge_parent_rtable(nf_bridge
->physindev
);
261 skb_dst_set_noref(skb
, &rt
->dst
);
263 skb
->dev
= nf_bridge
->physindev
;
264 nf_bridge_update_protocol(skb
);
265 nf_bridge_push_encap_header(skb
);
266 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
267 br_handle_frame_finish
, 1);
272 /* Obtain the correct destination MAC address, while preserving the original
273 * source MAC address. If we already know this address, we just copy it. If we
274 * don't, we use the neighbour framework to find out. In both cases, we make
275 * sure that br_handle_frame_finish() is called afterwards.
277 static int br_nf_pre_routing_finish_bridge(struct sk_buff
*skb
)
279 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
280 struct neighbour
*neigh
;
281 struct dst_entry
*dst
;
283 skb
->dev
= bridge_parent(skb
->dev
);
287 neigh
= dst_neigh_lookup_skb(dst
, skb
);
291 if (neigh
->hh
.hh_len
) {
292 neigh_hh_bridge(&neigh
->hh
, skb
);
293 skb
->dev
= nf_bridge
->physindev
;
294 ret
= br_handle_frame_finish(skb
);
296 /* the neighbour function below overwrites the complete
297 * MAC header, so we save the Ethernet source address and
300 skb_copy_from_linear_data_offset(skb
,
301 -(ETH_HLEN
-ETH_ALEN
),
302 skb
->nf_bridge
->data
,
304 /* tell br_dev_xmit to continue with forwarding */
305 nf_bridge
->mask
|= BRNF_BRIDGED_DNAT
;
306 /* FIXME Need to refragment */
307 ret
= neigh
->output(neigh
, skb
);
309 neigh_release(neigh
);
317 /* This requires some explaining. If DNAT has taken place,
318 * we will need to fix up the destination Ethernet address.
320 * There are two cases to consider:
321 * 1. The packet was DNAT'ed to a device in the same bridge
322 * port group as it was received on. We can still bridge
324 * 2. The packet was DNAT'ed to a different device, either
325 * a non-bridged device or another bridge port group.
326 * The packet will need to be routed.
328 * The correct way of distinguishing between these two cases is to
329 * call ip_route_input() and to look at skb->dst->dev, which is
330 * changed to the destination device if ip_route_input() succeeds.
332 * Let's first consider the case that ip_route_input() succeeds:
334 * If the output device equals the logical bridge device the packet
335 * came in on, we can consider this bridging. The corresponding MAC
336 * address will be obtained in br_nf_pre_routing_finish_bridge.
337 * Otherwise, the packet is considered to be routed and we just
338 * change the destination MAC address so that the packet will
339 * later be passed up to the IP stack to be routed. For a redirected
340 * packet, ip_route_input() will give back the localhost as output device,
341 * which differs from the bridge device.
343 * Let's now consider the case that ip_route_input() fails:
345 * This can be because the destination address is martian, in which case
346 * the packet will be dropped.
347 * If IP forwarding is disabled, ip_route_input() will fail, while
348 * ip_route_output_key() can return success. The source
349 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
350 * thinks we're handling a locally generated packet and won't care
351 * if IP forwarding is enabled. If the output device equals the logical bridge
352 * device, we proceed as if ip_route_input() succeeded. If it differs from the
353 * logical bridge port or if ip_route_output_key() fails we drop the packet.
355 static int br_nf_pre_routing_finish(struct sk_buff
*skb
)
357 struct net_device
*dev
= skb
->dev
;
358 struct iphdr
*iph
= ip_hdr(skb
);
359 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
364 frag_max_size
= IPCB(skb
)->frag_max_size
;
365 BR_INPUT_SKB_CB(skb
)->frag_max_size
= frag_max_size
;
367 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
368 skb
->pkt_type
= PACKET_OTHERHOST
;
369 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
371 nf_bridge
->mask
^= BRNF_NF_BRIDGE_PREROUTING
;
372 if (dnat_took_place(skb
)) {
373 if ((err
= ip_route_input(skb
, iph
->daddr
, iph
->saddr
, iph
->tos
, dev
))) {
374 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
376 /* If err equals -EHOSTUNREACH the error is due to a
377 * martian destination or due to the fact that
378 * forwarding is disabled. For most martian packets,
379 * ip_route_output_key() will fail. It won't fail for 2 types of
380 * martian destinations: loopback destinations and destination
381 * 0.0.0.0. In both cases the packet will be dropped because the
382 * destination is the loopback device and not the bridge. */
383 if (err
!= -EHOSTUNREACH
|| !in_dev
|| IN_DEV_FORWARD(in_dev
))
386 rt
= ip_route_output(dev_net(dev
), iph
->daddr
, 0,
387 RT_TOS(iph
->tos
), 0);
389 /* - Bridged-and-DNAT'ed traffic doesn't
390 * require ip_forwarding. */
391 if (rt
->dst
.dev
== dev
) {
392 skb_dst_set(skb
, &rt
->dst
);
401 if (skb_dst(skb
)->dev
== dev
) {
403 skb
->dev
= nf_bridge
->physindev
;
404 nf_bridge_update_protocol(skb
);
405 nf_bridge_push_encap_header(skb
);
406 NF_HOOK_THRESH(NFPROTO_BRIDGE
,
409 br_nf_pre_routing_finish_bridge
,
413 ether_addr_copy(eth_hdr(skb
)->h_dest
, dev
->dev_addr
);
414 skb
->pkt_type
= PACKET_HOST
;
417 rt
= bridge_parent_rtable(nf_bridge
->physindev
);
422 skb_dst_set_noref(skb
, &rt
->dst
);
425 skb
->dev
= nf_bridge
->physindev
;
426 nf_bridge_update_protocol(skb
);
427 nf_bridge_push_encap_header(skb
);
428 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
429 br_handle_frame_finish
, 1);
434 static struct net_device
*brnf_get_logical_dev(struct sk_buff
*skb
, const struct net_device
*dev
)
436 struct net_device
*vlan
, *br
;
438 br
= bridge_parent(dev
);
439 if (brnf_pass_vlan_indev
== 0 || !skb_vlan_tag_present(skb
))
442 vlan
= __vlan_find_dev_deep_rcu(br
, skb
->vlan_proto
,
443 skb_vlan_tag_get(skb
) & VLAN_VID_MASK
);
445 return vlan
? vlan
: br
;
448 /* Some common code for IPv4/IPv6 */
449 static struct net_device
*setup_pre_routing(struct sk_buff
*skb
)
451 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
453 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
454 skb
->pkt_type
= PACKET_HOST
;
455 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
458 nf_bridge
->mask
|= BRNF_NF_BRIDGE_PREROUTING
;
459 nf_bridge
->physindev
= skb
->dev
;
460 skb
->dev
= brnf_get_logical_dev(skb
, skb
->dev
);
461 if (skb
->protocol
== htons(ETH_P_8021Q
))
462 nf_bridge
->mask
|= BRNF_8021Q
;
463 else if (skb
->protocol
== htons(ETH_P_PPP_SES
))
464 nf_bridge
->mask
|= BRNF_PPPoE
;
466 /* Must drop socket now because of tproxy. */
471 /* We only check the length. A bridge shouldn't do any hop-by-hop stuff anyway */
472 static int check_hbh_len(struct sk_buff
*skb
)
474 unsigned char *raw
= (u8
*)(ipv6_hdr(skb
) + 1);
476 const unsigned char *nh
= skb_network_header(skb
);
478 int len
= (raw
[1] + 1) << 3;
480 if ((raw
+ len
) - skb
->data
> skb_headlen(skb
))
487 int optlen
= nh
[off
+ 1] + 2;
498 if (nh
[off
+ 1] != 4 || (off
& 3) != 2)
500 pkt_len
= ntohl(*(__be32
*) (nh
+ off
+ 2));
501 if (pkt_len
<= IPV6_MAXPLEN
||
502 ipv6_hdr(skb
)->payload_len
)
504 if (pkt_len
> skb
->len
- sizeof(struct ipv6hdr
))
506 if (pskb_trim_rcsum(skb
,
507 pkt_len
+ sizeof(struct ipv6hdr
)))
509 nh
= skb_network_header(skb
);
526 /* Replicate the checks that IPv6 does on packet reception and pass the packet
527 * to ip6tables, which doesn't support NAT, so things are fairly simple. */
528 static unsigned int br_nf_pre_routing_ipv6(const struct nf_hook_ops
*ops
,
530 const struct net_device
*in
,
531 const struct net_device
*out
,
532 int (*okfn
)(struct sk_buff
*))
534 const struct ipv6hdr
*hdr
;
537 if (skb
->len
< sizeof(struct ipv6hdr
))
540 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
545 if (hdr
->version
!= 6)
548 pkt_len
= ntohs(hdr
->payload_len
);
550 if (pkt_len
|| hdr
->nexthdr
!= NEXTHDR_HOP
) {
551 if (pkt_len
+ sizeof(struct ipv6hdr
) > skb
->len
)
553 if (pskb_trim_rcsum(skb
, pkt_len
+ sizeof(struct ipv6hdr
)))
556 if (hdr
->nexthdr
== NEXTHDR_HOP
&& check_hbh_len(skb
))
559 nf_bridge_put(skb
->nf_bridge
);
560 if (!nf_bridge_alloc(skb
))
562 if (!setup_pre_routing(skb
))
565 skb
->protocol
= htons(ETH_P_IPV6
);
566 NF_HOOK(NFPROTO_IPV6
, NF_INET_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
567 br_nf_pre_routing_finish_ipv6
);
572 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
573 * Replicate the checks that IPv4 does on packet reception.
574 * Set skb->dev to the bridge device (i.e. parent of the
575 * receiving device) to make netfilter happy, the REDIRECT
576 * target in particular. Save the original destination IP
577 * address to be able to detect DNAT afterwards. */
578 static unsigned int br_nf_pre_routing(const struct nf_hook_ops
*ops
,
580 const struct net_device
*in
,
581 const struct net_device
*out
,
582 int (*okfn
)(struct sk_buff
*))
584 struct net_bridge_port
*p
;
585 struct net_bridge
*br
;
586 __u32 len
= nf_bridge_encap_header_len(skb
);
588 if (unlikely(!pskb_may_pull(skb
, len
)))
591 p
= br_port_get_rcu(in
);
596 if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
)) {
597 if (!brnf_call_ip6tables
&& !br
->nf_call_ip6tables
)
600 nf_bridge_pull_encap_header_rcsum(skb
);
601 return br_nf_pre_routing_ipv6(ops
, skb
, in
, out
, okfn
);
604 if (!brnf_call_iptables
&& !br
->nf_call_iptables
)
607 if (!IS_IP(skb
) && !IS_VLAN_IP(skb
) && !IS_PPPOE_IP(skb
))
610 nf_bridge_pull_encap_header_rcsum(skb
);
612 if (br_parse_ip_options(skb
))
615 nf_bridge_put(skb
->nf_bridge
);
616 if (!nf_bridge_alloc(skb
))
618 if (!setup_pre_routing(skb
))
620 store_orig_dstaddr(skb
);
621 skb
->protocol
= htons(ETH_P_IP
);
623 NF_HOOK(NFPROTO_IPV4
, NF_INET_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
624 br_nf_pre_routing_finish
);
630 /* PF_BRIDGE/LOCAL_IN ************************************************/
631 /* The packet is locally destined, which requires a real
632 * dst_entry, so detach the fake one. On the way up, the
633 * packet would pass through PRE_ROUTING again (which already
634 * took place when the packet entered the bridge), but we
635 * register an IPv4 PRE_ROUTING 'sabotage' hook that will
636 * prevent this from happening. */
637 static unsigned int br_nf_local_in(const struct nf_hook_ops
*ops
,
639 const struct net_device
*in
,
640 const struct net_device
*out
,
641 int (*okfn
)(struct sk_buff
*))
643 br_drop_fake_rtable(skb
);
647 /* PF_BRIDGE/FORWARD *************************************************/
648 static int br_nf_forward_finish(struct sk_buff
*skb
)
650 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
651 struct net_device
*in
;
653 if (!IS_ARP(skb
) && !IS_VLAN_ARP(skb
)) {
654 in
= nf_bridge
->physindev
;
655 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
656 skb
->pkt_type
= PACKET_OTHERHOST
;
657 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
659 nf_bridge_update_protocol(skb
);
661 in
= *((struct net_device
**)(skb
->cb
));
663 nf_bridge_push_encap_header(skb
);
665 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_FORWARD
, skb
, in
,
666 skb
->dev
, br_forward_finish
, 1);
671 /* This is the 'purely bridged' case. For IP, we pass the packet to
672 * netfilter with indev and outdev set to the bridge device,
673 * but we are still able to filter on the 'real' indev/outdev
674 * because of the physdev module. For ARP, indev and outdev are the
676 static unsigned int br_nf_forward_ip(const struct nf_hook_ops
*ops
,
678 const struct net_device
*in
,
679 const struct net_device
*out
,
680 int (*okfn
)(struct sk_buff
*))
682 struct nf_bridge_info
*nf_bridge
;
683 struct net_device
*parent
;
689 /* Need exclusive nf_bridge_info since we might have multiple
690 * different physoutdevs. */
691 if (!nf_bridge_unshare(skb
))
694 parent
= bridge_parent(out
);
698 if (IS_IP(skb
) || IS_VLAN_IP(skb
) || IS_PPPOE_IP(skb
))
700 else if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
))
705 nf_bridge_pull_encap_header(skb
);
707 nf_bridge
= skb
->nf_bridge
;
708 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
709 skb
->pkt_type
= PACKET_HOST
;
710 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
713 if (pf
== NFPROTO_IPV4
&& br_parse_ip_options(skb
))
716 /* The physdev module checks on this */
717 nf_bridge
->mask
|= BRNF_BRIDGED
;
718 nf_bridge
->physoutdev
= skb
->dev
;
719 if (pf
== NFPROTO_IPV4
)
720 skb
->protocol
= htons(ETH_P_IP
);
722 skb
->protocol
= htons(ETH_P_IPV6
);
724 NF_HOOK(pf
, NF_INET_FORWARD
, skb
, brnf_get_logical_dev(skb
, in
), parent
,
725 br_nf_forward_finish
);
730 static unsigned int br_nf_forward_arp(const struct nf_hook_ops
*ops
,
732 const struct net_device
*in
,
733 const struct net_device
*out
,
734 int (*okfn
)(struct sk_buff
*))
736 struct net_bridge_port
*p
;
737 struct net_bridge
*br
;
738 struct net_device
**d
= (struct net_device
**)(skb
->cb
);
740 p
= br_port_get_rcu(out
);
745 if (!brnf_call_arptables
&& !br
->nf_call_arptables
)
749 if (!IS_VLAN_ARP(skb
))
751 nf_bridge_pull_encap_header(skb
);
754 if (arp_hdr(skb
)->ar_pln
!= 4) {
755 if (IS_VLAN_ARP(skb
))
756 nf_bridge_push_encap_header(skb
);
759 *d
= (struct net_device
*)in
;
760 NF_HOOK(NFPROTO_ARP
, NF_ARP_FORWARD
, skb
, (struct net_device
*)in
,
761 (struct net_device
*)out
, br_nf_forward_finish
);
766 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4)
767 static int br_nf_dev_queue_xmit(struct sk_buff
*skb
)
772 /* This is wrong! We should preserve the original fragment
773 * boundaries by preserving frag_list rather than refragmenting.
775 if (skb
->protocol
== htons(ETH_P_IP
) &&
776 skb
->len
+ nf_bridge_mtu_reduction(skb
) > skb
->dev
->mtu
&&
778 frag_max_size
= BR_INPUT_SKB_CB(skb
)->frag_max_size
;
779 if (br_parse_ip_options(skb
))
780 /* Drop invalid packet */
782 IPCB(skb
)->frag_max_size
= frag_max_size
;
783 ret
= ip_fragment(skb
, br_dev_queue_push_xmit
);
785 ret
= br_dev_queue_push_xmit(skb
);
790 static int br_nf_dev_queue_xmit(struct sk_buff
*skb
)
792 return br_dev_queue_push_xmit(skb
);
796 /* PF_BRIDGE/POST_ROUTING ********************************************/
797 static unsigned int br_nf_post_routing(const struct nf_hook_ops
*ops
,
799 const struct net_device
*in
,
800 const struct net_device
*out
,
801 int (*okfn
)(struct sk_buff
*))
803 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
804 struct net_device
*realoutdev
= bridge_parent(skb
->dev
);
807 if (!nf_bridge
|| !(nf_bridge
->mask
& BRNF_BRIDGED
))
813 if (IS_IP(skb
) || IS_VLAN_IP(skb
) || IS_PPPOE_IP(skb
))
815 else if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
))
820 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
821 * about the value of skb->pkt_type. */
822 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
823 skb
->pkt_type
= PACKET_HOST
;
824 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
827 nf_bridge_pull_encap_header(skb
);
828 nf_bridge_save_header(skb
);
829 if (pf
== NFPROTO_IPV4
)
830 skb
->protocol
= htons(ETH_P_IP
);
832 skb
->protocol
= htons(ETH_P_IPV6
);
834 NF_HOOK(pf
, NF_INET_POST_ROUTING
, skb
, NULL
, realoutdev
,
835 br_nf_dev_queue_xmit
);
840 /* IP/SABOTAGE *****************************************************/
841 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
842 * for the second time. */
843 static unsigned int ip_sabotage_in(const struct nf_hook_ops
*ops
,
845 const struct net_device
*in
,
846 const struct net_device
*out
,
847 int (*okfn
)(struct sk_buff
*))
849 if (skb
->nf_bridge
&&
850 !(skb
->nf_bridge
->mask
& BRNF_NF_BRIDGE_PREROUTING
)) {
857 void br_netfilter_enable(void)
860 EXPORT_SYMBOL_GPL(br_netfilter_enable
);
862 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
863 * br_dev_queue_push_xmit is called afterwards */
864 static struct nf_hook_ops br_nf_ops
[] __read_mostly
= {
866 .hook
= br_nf_pre_routing
,
867 .owner
= THIS_MODULE
,
868 .pf
= NFPROTO_BRIDGE
,
869 .hooknum
= NF_BR_PRE_ROUTING
,
870 .priority
= NF_BR_PRI_BRNF
,
873 .hook
= br_nf_local_in
,
874 .owner
= THIS_MODULE
,
875 .pf
= NFPROTO_BRIDGE
,
876 .hooknum
= NF_BR_LOCAL_IN
,
877 .priority
= NF_BR_PRI_BRNF
,
880 .hook
= br_nf_forward_ip
,
881 .owner
= THIS_MODULE
,
882 .pf
= NFPROTO_BRIDGE
,
883 .hooknum
= NF_BR_FORWARD
,
884 .priority
= NF_BR_PRI_BRNF
- 1,
887 .hook
= br_nf_forward_arp
,
888 .owner
= THIS_MODULE
,
889 .pf
= NFPROTO_BRIDGE
,
890 .hooknum
= NF_BR_FORWARD
,
891 .priority
= NF_BR_PRI_BRNF
,
894 .hook
= br_nf_post_routing
,
895 .owner
= THIS_MODULE
,
896 .pf
= NFPROTO_BRIDGE
,
897 .hooknum
= NF_BR_POST_ROUTING
,
898 .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
,
908 .hook
= ip_sabotage_in
,
909 .owner
= THIS_MODULE
,
911 .hooknum
= NF_INET_PRE_ROUTING
,
912 .priority
= NF_IP6_PRI_FIRST
,
918 int brnf_sysctl_call_tables(struct ctl_table
*ctl
, int write
,
919 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
923 ret
= proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
925 if (write
&& *(int *)(ctl
->data
))
926 *(int *)(ctl
->data
) = 1;
930 static struct ctl_table brnf_table
[] = {
932 .procname
= "bridge-nf-call-arptables",
933 .data
= &brnf_call_arptables
,
934 .maxlen
= sizeof(int),
936 .proc_handler
= brnf_sysctl_call_tables
,
939 .procname
= "bridge-nf-call-iptables",
940 .data
= &brnf_call_iptables
,
941 .maxlen
= sizeof(int),
943 .proc_handler
= brnf_sysctl_call_tables
,
946 .procname
= "bridge-nf-call-ip6tables",
947 .data
= &brnf_call_ip6tables
,
948 .maxlen
= sizeof(int),
950 .proc_handler
= brnf_sysctl_call_tables
,
953 .procname
= "bridge-nf-filter-vlan-tagged",
954 .data
= &brnf_filter_vlan_tagged
,
955 .maxlen
= sizeof(int),
957 .proc_handler
= brnf_sysctl_call_tables
,
960 .procname
= "bridge-nf-filter-pppoe-tagged",
961 .data
= &brnf_filter_pppoe_tagged
,
962 .maxlen
= sizeof(int),
964 .proc_handler
= brnf_sysctl_call_tables
,
967 .procname
= "bridge-nf-pass-vlan-input-dev",
968 .data
= &brnf_pass_vlan_indev
,
969 .maxlen
= sizeof(int),
971 .proc_handler
= brnf_sysctl_call_tables
,
977 static int __init
br_netfilter_init(void)
981 ret
= nf_register_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
986 brnf_sysctl_header
= register_net_sysctl(&init_net
, "net/bridge", brnf_table
);
987 if (brnf_sysctl_header
== NULL
) {
989 "br_netfilter: can't register to sysctl.\n");
990 nf_unregister_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
994 printk(KERN_NOTICE
"Bridge firewalling registered\n");
998 static void __exit
br_netfilter_fini(void)
1000 nf_unregister_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
1001 #ifdef CONFIG_SYSCTL
1002 unregister_net_sysctl_table(brnf_sysctl_header
);
1006 module_init(br_netfilter_init
);
1007 module_exit(br_netfilter_fini
);
1009 MODULE_LICENSE("GPL");
1010 MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
1011 MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
1012 MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");