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 static void nf_bridge_update_protocol(struct sk_buff
*skb
)
244 if (skb
->nf_bridge
->mask
& BRNF_8021Q
)
245 skb
->protocol
= htons(ETH_P_8021Q
);
246 else if (skb
->nf_bridge
->mask
& BRNF_PPPoE
)
247 skb
->protocol
= htons(ETH_P_PPP_SES
);
250 /* PF_BRIDGE/PRE_ROUTING *********************************************/
251 /* Undo the changes made for ip6tables PREROUTING and continue the
252 * bridge PRE_ROUTING hook. */
253 static int br_nf_pre_routing_finish_ipv6(struct sk_buff
*skb
)
255 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
258 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
259 skb
->pkt_type
= PACKET_OTHERHOST
;
260 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
262 nf_bridge
->mask
^= BRNF_NF_BRIDGE_PREROUTING
;
264 rt
= bridge_parent_rtable(nf_bridge
->physindev
);
269 skb_dst_set_noref(skb
, &rt
->dst
);
271 skb
->dev
= nf_bridge
->physindev
;
272 nf_bridge_update_protocol(skb
);
273 nf_bridge_push_encap_header(skb
);
274 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
275 br_handle_frame_finish
, 1);
280 /* Obtain the correct destination MAC address, while preserving the original
281 * source MAC address. If we already know this address, we just copy it. If we
282 * don't, we use the neighbour framework to find out. In both cases, we make
283 * sure that br_handle_frame_finish() is called afterwards.
285 static int br_nf_pre_routing_finish_bridge(struct sk_buff
*skb
)
287 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
288 struct neighbour
*neigh
;
289 struct dst_entry
*dst
;
291 skb
->dev
= bridge_parent(skb
->dev
);
295 neigh
= dst_neigh_lookup_skb(dst
, skb
);
299 if (neigh
->hh
.hh_len
) {
300 neigh_hh_bridge(&neigh
->hh
, skb
);
301 skb
->dev
= nf_bridge
->physindev
;
302 ret
= br_handle_frame_finish(skb
);
304 /* the neighbour function below overwrites the complete
305 * MAC header, so we save the Ethernet source address and
308 skb_copy_from_linear_data_offset(skb
,
309 -(ETH_HLEN
-ETH_ALEN
),
310 skb
->nf_bridge
->data
,
312 /* tell br_dev_xmit to continue with forwarding */
313 nf_bridge
->mask
|= BRNF_BRIDGED_DNAT
;
314 /* FIXME Need to refragment */
315 ret
= neigh
->output(neigh
, skb
);
317 neigh_release(neigh
);
325 /* This requires some explaining. If DNAT has taken place,
326 * we will need to fix up the destination Ethernet address.
328 * There are two cases to consider:
329 * 1. The packet was DNAT'ed to a device in the same bridge
330 * port group as it was received on. We can still bridge
332 * 2. The packet was DNAT'ed to a different device, either
333 * a non-bridged device or another bridge port group.
334 * The packet will need to be routed.
336 * The correct way of distinguishing between these two cases is to
337 * call ip_route_input() and to look at skb->dst->dev, which is
338 * changed to the destination device if ip_route_input() succeeds.
340 * Let's first consider the case that ip_route_input() succeeds:
342 * If the output device equals the logical bridge device the packet
343 * came in on, we can consider this bridging. The corresponding MAC
344 * address will be obtained in br_nf_pre_routing_finish_bridge.
345 * Otherwise, the packet is considered to be routed and we just
346 * change the destination MAC address so that the packet will
347 * later be passed up to the IP stack to be routed. For a redirected
348 * packet, ip_route_input() will give back the localhost as output device,
349 * which differs from the bridge device.
351 * Let's now consider the case that ip_route_input() fails:
353 * This can be because the destination address is martian, in which case
354 * the packet will be dropped.
355 * If IP forwarding is disabled, ip_route_input() will fail, while
356 * ip_route_output_key() can return success. The source
357 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
358 * thinks we're handling a locally generated packet and won't care
359 * if IP forwarding is enabled. If the output device equals the logical bridge
360 * device, we proceed as if ip_route_input() succeeded. If it differs from the
361 * logical bridge port or if ip_route_output_key() fails we drop the packet.
363 static int br_nf_pre_routing_finish(struct sk_buff
*skb
)
365 struct net_device
*dev
= skb
->dev
;
366 struct iphdr
*iph
= ip_hdr(skb
);
367 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
372 frag_max_size
= IPCB(skb
)->frag_max_size
;
373 BR_INPUT_SKB_CB(skb
)->frag_max_size
= frag_max_size
;
375 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
376 skb
->pkt_type
= PACKET_OTHERHOST
;
377 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
379 nf_bridge
->mask
^= BRNF_NF_BRIDGE_PREROUTING
;
380 if (dnat_took_place(skb
)) {
381 if ((err
= ip_route_input(skb
, iph
->daddr
, iph
->saddr
, iph
->tos
, dev
))) {
382 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
384 /* If err equals -EHOSTUNREACH the error is due to a
385 * martian destination or due to the fact that
386 * forwarding is disabled. For most martian packets,
387 * ip_route_output_key() will fail. It won't fail for 2 types of
388 * martian destinations: loopback destinations and destination
389 * 0.0.0.0. In both cases the packet will be dropped because the
390 * destination is the loopback device and not the bridge. */
391 if (err
!= -EHOSTUNREACH
|| !in_dev
|| IN_DEV_FORWARD(in_dev
))
394 rt
= ip_route_output(dev_net(dev
), iph
->daddr
, 0,
395 RT_TOS(iph
->tos
), 0);
397 /* - Bridged-and-DNAT'ed traffic doesn't
398 * require ip_forwarding. */
399 if (rt
->dst
.dev
== dev
) {
400 skb_dst_set(skb
, &rt
->dst
);
409 if (skb_dst(skb
)->dev
== dev
) {
411 skb
->dev
= nf_bridge
->physindev
;
412 nf_bridge_update_protocol(skb
);
413 nf_bridge_push_encap_header(skb
);
414 NF_HOOK_THRESH(NFPROTO_BRIDGE
,
417 br_nf_pre_routing_finish_bridge
,
421 ether_addr_copy(eth_hdr(skb
)->h_dest
, dev
->dev_addr
);
422 skb
->pkt_type
= PACKET_HOST
;
425 rt
= bridge_parent_rtable(nf_bridge
->physindev
);
430 skb_dst_set_noref(skb
, &rt
->dst
);
433 skb
->dev
= nf_bridge
->physindev
;
434 nf_bridge_update_protocol(skb
);
435 nf_bridge_push_encap_header(skb
);
436 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
437 br_handle_frame_finish
, 1);
442 static struct net_device
*brnf_get_logical_dev(struct sk_buff
*skb
, const struct net_device
*dev
)
444 struct net_device
*vlan
, *br
;
446 br
= bridge_parent(dev
);
447 if (brnf_pass_vlan_indev
== 0 || !skb_vlan_tag_present(skb
))
450 vlan
= __vlan_find_dev_deep_rcu(br
, skb
->vlan_proto
,
451 skb_vlan_tag_get(skb
) & VLAN_VID_MASK
);
453 return vlan
? vlan
: br
;
456 /* Some common code for IPv4/IPv6 */
457 static struct net_device
*setup_pre_routing(struct sk_buff
*skb
)
459 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
461 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
462 skb
->pkt_type
= PACKET_HOST
;
463 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
466 nf_bridge
->mask
|= BRNF_NF_BRIDGE_PREROUTING
;
467 nf_bridge
->physindev
= skb
->dev
;
468 skb
->dev
= brnf_get_logical_dev(skb
, skb
->dev
);
469 if (skb
->protocol
== htons(ETH_P_8021Q
))
470 nf_bridge
->mask
|= BRNF_8021Q
;
471 else if (skb
->protocol
== htons(ETH_P_PPP_SES
))
472 nf_bridge
->mask
|= BRNF_PPPoE
;
474 /* Must drop socket now because of tproxy. */
479 /* We only check the length. A bridge shouldn't do any hop-by-hop stuff anyway */
480 static int check_hbh_len(struct sk_buff
*skb
)
482 unsigned char *raw
= (u8
*)(ipv6_hdr(skb
) + 1);
484 const unsigned char *nh
= skb_network_header(skb
);
486 int len
= (raw
[1] + 1) << 3;
488 if ((raw
+ len
) - skb
->data
> skb_headlen(skb
))
495 int optlen
= nh
[off
+ 1] + 2;
506 if (nh
[off
+ 1] != 4 || (off
& 3) != 2)
508 pkt_len
= ntohl(*(__be32
*) (nh
+ off
+ 2));
509 if (pkt_len
<= IPV6_MAXPLEN
||
510 ipv6_hdr(skb
)->payload_len
)
512 if (pkt_len
> skb
->len
- sizeof(struct ipv6hdr
))
514 if (pskb_trim_rcsum(skb
,
515 pkt_len
+ sizeof(struct ipv6hdr
)))
517 nh
= skb_network_header(skb
);
534 /* Replicate the checks that IPv6 does on packet reception and pass the packet
535 * to ip6tables, which doesn't support NAT, so things are fairly simple. */
536 static unsigned int br_nf_pre_routing_ipv6(const struct nf_hook_ops
*ops
,
538 const struct net_device
*in
,
539 const struct net_device
*out
,
540 int (*okfn
)(struct sk_buff
*))
542 const struct ipv6hdr
*hdr
;
545 if (skb
->len
< sizeof(struct ipv6hdr
))
548 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
553 if (hdr
->version
!= 6)
556 pkt_len
= ntohs(hdr
->payload_len
);
558 if (pkt_len
|| hdr
->nexthdr
!= NEXTHDR_HOP
) {
559 if (pkt_len
+ sizeof(struct ipv6hdr
) > skb
->len
)
561 if (pskb_trim_rcsum(skb
, pkt_len
+ sizeof(struct ipv6hdr
)))
564 if (hdr
->nexthdr
== NEXTHDR_HOP
&& check_hbh_len(skb
))
567 nf_bridge_put(skb
->nf_bridge
);
568 if (!nf_bridge_alloc(skb
))
570 if (!setup_pre_routing(skb
))
573 skb
->protocol
= htons(ETH_P_IPV6
);
574 NF_HOOK(NFPROTO_IPV6
, NF_INET_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
575 br_nf_pre_routing_finish_ipv6
);
580 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
581 * Replicate the checks that IPv4 does on packet reception.
582 * Set skb->dev to the bridge device (i.e. parent of the
583 * receiving device) to make netfilter happy, the REDIRECT
584 * target in particular. Save the original destination IP
585 * address to be able to detect DNAT afterwards. */
586 static unsigned int br_nf_pre_routing(const struct nf_hook_ops
*ops
,
588 const struct net_device
*in
,
589 const struct net_device
*out
,
590 int (*okfn
)(struct sk_buff
*))
592 struct net_bridge_port
*p
;
593 struct net_bridge
*br
;
594 __u32 len
= nf_bridge_encap_header_len(skb
);
596 if (unlikely(!pskb_may_pull(skb
, len
)))
599 p
= br_port_get_rcu(in
);
604 if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
)) {
605 if (!brnf_call_ip6tables
&& !br
->nf_call_ip6tables
)
608 nf_bridge_pull_encap_header_rcsum(skb
);
609 return br_nf_pre_routing_ipv6(ops
, skb
, in
, out
, okfn
);
612 if (!brnf_call_iptables
&& !br
->nf_call_iptables
)
615 if (!IS_IP(skb
) && !IS_VLAN_IP(skb
) && !IS_PPPOE_IP(skb
))
618 nf_bridge_pull_encap_header_rcsum(skb
);
620 if (br_parse_ip_options(skb
))
623 nf_bridge_put(skb
->nf_bridge
);
624 if (!nf_bridge_alloc(skb
))
626 if (!setup_pre_routing(skb
))
628 store_orig_dstaddr(skb
);
629 skb
->protocol
= htons(ETH_P_IP
);
631 NF_HOOK(NFPROTO_IPV4
, NF_INET_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
632 br_nf_pre_routing_finish
);
638 /* PF_BRIDGE/LOCAL_IN ************************************************/
639 /* The packet is locally destined, which requires a real
640 * dst_entry, so detach the fake one. On the way up, the
641 * packet would pass through PRE_ROUTING again (which already
642 * took place when the packet entered the bridge), but we
643 * register an IPv4 PRE_ROUTING 'sabotage' hook that will
644 * prevent this from happening. */
645 static unsigned int br_nf_local_in(const struct nf_hook_ops
*ops
,
647 const struct net_device
*in
,
648 const struct net_device
*out
,
649 int (*okfn
)(struct sk_buff
*))
651 br_drop_fake_rtable(skb
);
655 /* PF_BRIDGE/FORWARD *************************************************/
656 static int br_nf_forward_finish(struct sk_buff
*skb
)
658 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
659 struct net_device
*in
;
661 if (!IS_ARP(skb
) && !IS_VLAN_ARP(skb
)) {
662 in
= nf_bridge
->physindev
;
663 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
664 skb
->pkt_type
= PACKET_OTHERHOST
;
665 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
667 nf_bridge_update_protocol(skb
);
669 in
= *((struct net_device
**)(skb
->cb
));
671 nf_bridge_push_encap_header(skb
);
673 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_FORWARD
, skb
, in
,
674 skb
->dev
, br_forward_finish
, 1);
679 /* This is the 'purely bridged' case. For IP, we pass the packet to
680 * netfilter with indev and outdev set to the bridge device,
681 * but we are still able to filter on the 'real' indev/outdev
682 * because of the physdev module. For ARP, indev and outdev are the
684 static unsigned int br_nf_forward_ip(const struct nf_hook_ops
*ops
,
686 const struct net_device
*in
,
687 const struct net_device
*out
,
688 int (*okfn
)(struct sk_buff
*))
690 struct nf_bridge_info
*nf_bridge
;
691 struct net_device
*parent
;
697 /* Need exclusive nf_bridge_info since we might have multiple
698 * different physoutdevs. */
699 if (!nf_bridge_unshare(skb
))
702 parent
= bridge_parent(out
);
706 if (IS_IP(skb
) || IS_VLAN_IP(skb
) || IS_PPPOE_IP(skb
))
708 else if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
))
713 nf_bridge_pull_encap_header(skb
);
715 nf_bridge
= skb
->nf_bridge
;
716 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
717 skb
->pkt_type
= PACKET_HOST
;
718 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
721 if (pf
== NFPROTO_IPV4
&& br_parse_ip_options(skb
))
724 /* The physdev module checks on this */
725 nf_bridge
->mask
|= BRNF_BRIDGED
;
726 nf_bridge
->physoutdev
= skb
->dev
;
727 if (pf
== NFPROTO_IPV4
)
728 skb
->protocol
= htons(ETH_P_IP
);
730 skb
->protocol
= htons(ETH_P_IPV6
);
732 NF_HOOK(pf
, NF_INET_FORWARD
, skb
, brnf_get_logical_dev(skb
, in
), parent
,
733 br_nf_forward_finish
);
738 static unsigned int br_nf_forward_arp(const struct nf_hook_ops
*ops
,
740 const struct net_device
*in
,
741 const struct net_device
*out
,
742 int (*okfn
)(struct sk_buff
*))
744 struct net_bridge_port
*p
;
745 struct net_bridge
*br
;
746 struct net_device
**d
= (struct net_device
**)(skb
->cb
);
748 p
= br_port_get_rcu(out
);
753 if (!brnf_call_arptables
&& !br
->nf_call_arptables
)
757 if (!IS_VLAN_ARP(skb
))
759 nf_bridge_pull_encap_header(skb
);
762 if (arp_hdr(skb
)->ar_pln
!= 4) {
763 if (IS_VLAN_ARP(skb
))
764 nf_bridge_push_encap_header(skb
);
767 *d
= (struct net_device
*)in
;
768 NF_HOOK(NFPROTO_ARP
, NF_ARP_FORWARD
, skb
, (struct net_device
*)in
,
769 (struct net_device
*)out
, br_nf_forward_finish
);
774 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4)
775 static bool nf_bridge_copy_header(struct sk_buff
*skb
)
778 unsigned int header_size
;
780 nf_bridge_update_protocol(skb
);
781 header_size
= ETH_HLEN
+ nf_bridge_encap_header_len(skb
);
782 err
= skb_cow_head(skb
, header_size
);
786 skb_copy_to_linear_data_offset(skb
, -header_size
,
787 skb
->nf_bridge
->data
, header_size
);
788 __skb_push(skb
, nf_bridge_encap_header_len(skb
));
792 static int br_nf_push_frag_xmit(struct sk_buff
*skb
)
794 if (!nf_bridge_copy_header(skb
)) {
799 return br_dev_queue_push_xmit(skb
);
802 static int br_nf_dev_queue_xmit(struct sk_buff
*skb
)
806 unsigned int mtu_reserved
;
808 if (skb_is_gso(skb
) || skb
->protocol
!= htons(ETH_P_IP
))
809 return br_dev_queue_push_xmit(skb
);
811 mtu_reserved
= nf_bridge_mtu_reduction(skb
);
812 /* This is wrong! We should preserve the original fragment
813 * boundaries by preserving frag_list rather than refragmenting.
815 if (skb
->len
+ mtu_reserved
> skb
->dev
->mtu
) {
816 frag_max_size
= BR_INPUT_SKB_CB(skb
)->frag_max_size
;
817 if (br_parse_ip_options(skb
))
818 /* Drop invalid packet */
820 IPCB(skb
)->frag_max_size
= frag_max_size
;
821 ret
= ip_fragment(skb
, br_nf_push_frag_xmit
);
823 ret
= br_dev_queue_push_xmit(skb
);
828 static int br_nf_dev_queue_xmit(struct sk_buff
*skb
)
830 return br_dev_queue_push_xmit(skb
);
834 /* PF_BRIDGE/POST_ROUTING ********************************************/
835 static unsigned int br_nf_post_routing(const struct nf_hook_ops
*ops
,
837 const struct net_device
*in
,
838 const struct net_device
*out
,
839 int (*okfn
)(struct sk_buff
*))
841 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
842 struct net_device
*realoutdev
= bridge_parent(skb
->dev
);
845 if (!nf_bridge
|| !(nf_bridge
->mask
& BRNF_BRIDGED
))
851 if (IS_IP(skb
) || IS_VLAN_IP(skb
) || IS_PPPOE_IP(skb
))
853 else if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
))
858 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
859 * about the value of skb->pkt_type. */
860 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
861 skb
->pkt_type
= PACKET_HOST
;
862 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
865 nf_bridge_pull_encap_header(skb
);
866 nf_bridge_save_header(skb
);
867 if (pf
== NFPROTO_IPV4
)
868 skb
->protocol
= htons(ETH_P_IP
);
870 skb
->protocol
= htons(ETH_P_IPV6
);
872 NF_HOOK(pf
, NF_INET_POST_ROUTING
, skb
, NULL
, realoutdev
,
873 br_nf_dev_queue_xmit
);
878 /* IP/SABOTAGE *****************************************************/
879 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
880 * for the second time. */
881 static unsigned int ip_sabotage_in(const struct nf_hook_ops
*ops
,
883 const struct net_device
*in
,
884 const struct net_device
*out
,
885 int (*okfn
)(struct sk_buff
*))
887 if (skb
->nf_bridge
&&
888 !(skb
->nf_bridge
->mask
& BRNF_NF_BRIDGE_PREROUTING
)) {
895 /* This is called when br_netfilter has called into iptables/netfilter,
896 * and DNAT has taken place on a bridge-forwarded packet.
898 * neigh->output has created a new MAC header, with local br0 MAC
901 * This restores the original MAC saddr of the bridged packet
902 * before invoking bridge forward logic to transmit the packet.
904 static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff
*skb
)
906 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
908 skb_pull(skb
, ETH_HLEN
);
909 nf_bridge
->mask
&= ~BRNF_BRIDGED_DNAT
;
911 skb_copy_to_linear_data_offset(skb
, -(ETH_HLEN
-ETH_ALEN
),
912 skb
->nf_bridge
->data
, ETH_HLEN
-ETH_ALEN
);
913 skb
->dev
= nf_bridge
->physindev
;
914 br_handle_frame_finish(skb
);
917 int br_nf_prerouting_finish_bridge(struct sk_buff
*skb
)
919 if (skb
->nf_bridge
&& (skb
->nf_bridge
->mask
& BRNF_BRIDGED_DNAT
)) {
920 br_nf_pre_routing_finish_bridge_slow(skb
);
925 EXPORT_SYMBOL_GPL(br_nf_prerouting_finish_bridge
);
927 void br_netfilter_enable(void)
930 EXPORT_SYMBOL_GPL(br_netfilter_enable
);
932 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
933 * br_dev_queue_push_xmit is called afterwards */
934 static struct nf_hook_ops br_nf_ops
[] __read_mostly
= {
936 .hook
= br_nf_pre_routing
,
937 .owner
= THIS_MODULE
,
938 .pf
= NFPROTO_BRIDGE
,
939 .hooknum
= NF_BR_PRE_ROUTING
,
940 .priority
= NF_BR_PRI_BRNF
,
943 .hook
= br_nf_local_in
,
944 .owner
= THIS_MODULE
,
945 .pf
= NFPROTO_BRIDGE
,
946 .hooknum
= NF_BR_LOCAL_IN
,
947 .priority
= NF_BR_PRI_BRNF
,
950 .hook
= br_nf_forward_ip
,
951 .owner
= THIS_MODULE
,
952 .pf
= NFPROTO_BRIDGE
,
953 .hooknum
= NF_BR_FORWARD
,
954 .priority
= NF_BR_PRI_BRNF
- 1,
957 .hook
= br_nf_forward_arp
,
958 .owner
= THIS_MODULE
,
959 .pf
= NFPROTO_BRIDGE
,
960 .hooknum
= NF_BR_FORWARD
,
961 .priority
= NF_BR_PRI_BRNF
,
964 .hook
= br_nf_post_routing
,
965 .owner
= THIS_MODULE
,
966 .pf
= NFPROTO_BRIDGE
,
967 .hooknum
= NF_BR_POST_ROUTING
,
968 .priority
= NF_BR_PRI_LAST
,
971 .hook
= ip_sabotage_in
,
972 .owner
= THIS_MODULE
,
974 .hooknum
= NF_INET_PRE_ROUTING
,
975 .priority
= NF_IP_PRI_FIRST
,
978 .hook
= ip_sabotage_in
,
979 .owner
= THIS_MODULE
,
981 .hooknum
= NF_INET_PRE_ROUTING
,
982 .priority
= NF_IP6_PRI_FIRST
,
988 int brnf_sysctl_call_tables(struct ctl_table
*ctl
, int write
,
989 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
993 ret
= proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
995 if (write
&& *(int *)(ctl
->data
))
996 *(int *)(ctl
->data
) = 1;
1000 static struct ctl_table brnf_table
[] = {
1002 .procname
= "bridge-nf-call-arptables",
1003 .data
= &brnf_call_arptables
,
1004 .maxlen
= sizeof(int),
1006 .proc_handler
= brnf_sysctl_call_tables
,
1009 .procname
= "bridge-nf-call-iptables",
1010 .data
= &brnf_call_iptables
,
1011 .maxlen
= sizeof(int),
1013 .proc_handler
= brnf_sysctl_call_tables
,
1016 .procname
= "bridge-nf-call-ip6tables",
1017 .data
= &brnf_call_ip6tables
,
1018 .maxlen
= sizeof(int),
1020 .proc_handler
= brnf_sysctl_call_tables
,
1023 .procname
= "bridge-nf-filter-vlan-tagged",
1024 .data
= &brnf_filter_vlan_tagged
,
1025 .maxlen
= sizeof(int),
1027 .proc_handler
= brnf_sysctl_call_tables
,
1030 .procname
= "bridge-nf-filter-pppoe-tagged",
1031 .data
= &brnf_filter_pppoe_tagged
,
1032 .maxlen
= sizeof(int),
1034 .proc_handler
= brnf_sysctl_call_tables
,
1037 .procname
= "bridge-nf-pass-vlan-input-dev",
1038 .data
= &brnf_pass_vlan_indev
,
1039 .maxlen
= sizeof(int),
1041 .proc_handler
= brnf_sysctl_call_tables
,
1047 static int __init
br_netfilter_init(void)
1051 ret
= nf_register_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
1055 #ifdef CONFIG_SYSCTL
1056 brnf_sysctl_header
= register_net_sysctl(&init_net
, "net/bridge", brnf_table
);
1057 if (brnf_sysctl_header
== NULL
) {
1059 "br_netfilter: can't register to sysctl.\n");
1060 nf_unregister_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
1064 printk(KERN_NOTICE
"Bridge firewalling registered\n");
1068 static void __exit
br_netfilter_fini(void)
1070 nf_unregister_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
1071 #ifdef CONFIG_SYSCTL
1072 unregister_net_sysctl_table(brnf_sysctl_header
);
1076 module_init(br_netfilter_init
);
1077 module_exit(br_netfilter_fini
);
1079 MODULE_LICENSE("GPL");
1080 MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
1081 MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
1082 MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");