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/addrconf.h>
38 #include <net/route.h>
39 #include <net/netfilter/br_netfilter.h>
41 #include <asm/uaccess.h>
42 #include "br_private.h"
44 #include <linux/sysctl.h>
48 static struct ctl_table_header
*brnf_sysctl_header
;
49 static int brnf_call_iptables __read_mostly
= 1;
50 static int brnf_call_ip6tables __read_mostly
= 1;
51 static int brnf_call_arptables __read_mostly
= 1;
52 static int brnf_filter_vlan_tagged __read_mostly
;
53 static int brnf_filter_pppoe_tagged __read_mostly
;
54 static int brnf_pass_vlan_indev __read_mostly
;
56 #define brnf_call_iptables 1
57 #define brnf_call_ip6tables 1
58 #define brnf_call_arptables 1
59 #define brnf_filter_vlan_tagged 0
60 #define brnf_filter_pppoe_tagged 0
61 #define brnf_pass_vlan_indev 0
65 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP))
67 #define IS_IPV6(skb) \
68 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6))
71 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP))
73 static inline __be16
vlan_proto(const struct sk_buff
*skb
)
75 if (skb_vlan_tag_present(skb
))
77 else if (skb
->protocol
== htons(ETH_P_8021Q
))
78 return vlan_eth_hdr(skb
)->h_vlan_encapsulated_proto
;
83 #define IS_VLAN_IP(skb) \
84 (vlan_proto(skb) == htons(ETH_P_IP) && \
85 brnf_filter_vlan_tagged)
87 #define IS_VLAN_IPV6(skb) \
88 (vlan_proto(skb) == htons(ETH_P_IPV6) && \
89 brnf_filter_vlan_tagged)
91 #define IS_VLAN_ARP(skb) \
92 (vlan_proto(skb) == htons(ETH_P_ARP) && \
93 brnf_filter_vlan_tagged)
95 static inline __be16
pppoe_proto(const struct sk_buff
*skb
)
97 return *((__be16
*)(skb_mac_header(skb
) + ETH_HLEN
+
98 sizeof(struct pppoe_hdr
)));
101 #define IS_PPPOE_IP(skb) \
102 (skb->protocol == htons(ETH_P_PPP_SES) && \
103 pppoe_proto(skb) == htons(PPP_IP) && \
104 brnf_filter_pppoe_tagged)
106 #define IS_PPPOE_IPV6(skb) \
107 (skb->protocol == htons(ETH_P_PPP_SES) && \
108 pppoe_proto(skb) == htons(PPP_IPV6) && \
109 brnf_filter_pppoe_tagged)
111 /* largest possible L2 header, see br_nf_dev_queue_xmit() */
112 #define NF_BRIDGE_MAX_MAC_HEADER_LENGTH (PPPOE_SES_HLEN + ETH_HLEN)
114 struct brnf_frag_data
{
115 char mac
[NF_BRIDGE_MAX_MAC_HEADER_LENGTH
];
122 static DEFINE_PER_CPU(struct brnf_frag_data
, brnf_frag_data_storage
);
124 static void nf_bridge_info_free(struct sk_buff
*skb
)
126 if (skb
->nf_bridge
) {
127 nf_bridge_put(skb
->nf_bridge
);
128 skb
->nf_bridge
= NULL
;
132 static inline struct net_device
*bridge_parent(const struct net_device
*dev
)
134 struct net_bridge_port
*port
;
136 port
= br_port_get_rcu(dev
);
137 return port
? port
->br
->dev
: NULL
;
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 unsigned int nf_bridge_encap_header_len(const struct sk_buff
*skb
)
159 switch (skb
->protocol
) {
160 case __cpu_to_be16(ETH_P_8021Q
):
162 case __cpu_to_be16(ETH_P_PPP_SES
):
163 return PPPOE_SES_HLEN
;
169 static inline void nf_bridge_pull_encap_header(struct sk_buff
*skb
)
171 unsigned int len
= nf_bridge_encap_header_len(skb
);
174 skb
->network_header
+= len
;
177 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff
*skb
)
179 unsigned int len
= nf_bridge_encap_header_len(skb
);
181 skb_pull_rcsum(skb
, len
);
182 skb
->network_header
+= len
;
185 /* When handing a packet over to the IP layer
186 * check whether we have a skb that is in the
190 static int br_validate_ipv4(struct net
*net
, struct sk_buff
*skb
)
192 const struct iphdr
*iph
;
195 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
200 /* Basic sanity checks */
201 if (iph
->ihl
< 5 || iph
->version
!= 4)
204 if (!pskb_may_pull(skb
, iph
->ihl
*4))
208 if (unlikely(ip_fast_csum((u8
*)iph
, iph
->ihl
)))
211 len
= ntohs(iph
->tot_len
);
212 if (skb
->len
< len
) {
213 IP_INC_STATS_BH(net
, IPSTATS_MIB_INTRUNCATEDPKTS
);
215 } else if (len
< (iph
->ihl
*4))
218 if (pskb_trim_rcsum(skb
, len
)) {
219 IP_INC_STATS_BH(net
, IPSTATS_MIB_INDISCARDS
);
223 memset(IPCB(skb
), 0, sizeof(struct inet_skb_parm
));
224 /* We should really parse IP options here but until
225 * somebody who actually uses IP options complains to
226 * us we'll just silently ignore the options because
232 IP_INC_STATS_BH(net
, IPSTATS_MIB_INHDRERRORS
);
237 void nf_bridge_update_protocol(struct sk_buff
*skb
)
239 switch (skb
->nf_bridge
->orig_proto
) {
240 case BRNF_PROTO_8021Q
:
241 skb
->protocol
= htons(ETH_P_8021Q
);
243 case BRNF_PROTO_PPPOE
:
244 skb
->protocol
= htons(ETH_P_PPP_SES
);
246 case BRNF_PROTO_UNCHANGED
:
251 /* Obtain the correct destination MAC address, while preserving the original
252 * source MAC address. If we already know this address, we just copy it. If we
253 * don't, we use the neighbour framework to find out. In both cases, we make
254 * sure that br_handle_frame_finish() is called afterwards.
256 int br_nf_pre_routing_finish_bridge(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
258 struct neighbour
*neigh
;
259 struct dst_entry
*dst
;
261 skb
->dev
= bridge_parent(skb
->dev
);
265 neigh
= dst_neigh_lookup_skb(dst
, skb
);
267 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
270 if (neigh
->hh
.hh_len
) {
271 neigh_hh_bridge(&neigh
->hh
, skb
);
272 skb
->dev
= nf_bridge
->physindev
;
273 ret
= br_handle_frame_finish(net
, sk
, skb
);
275 /* the neighbour function below overwrites the complete
276 * MAC header, so we save the Ethernet source address and
279 skb_copy_from_linear_data_offset(skb
,
280 -(ETH_HLEN
-ETH_ALEN
),
281 nf_bridge
->neigh_header
,
283 /* tell br_dev_xmit to continue with forwarding */
284 nf_bridge
->bridged_dnat
= 1;
285 /* FIXME Need to refragment */
286 ret
= neigh
->output(neigh
, skb
);
288 neigh_release(neigh
);
297 br_nf_ipv4_daddr_was_changed(const struct sk_buff
*skb
,
298 const struct nf_bridge_info
*nf_bridge
)
300 return ip_hdr(skb
)->daddr
!= nf_bridge
->ipv4_daddr
;
303 /* This requires some explaining. If DNAT has taken place,
304 * we will need to fix up the destination Ethernet address.
305 * This is also true when SNAT takes place (for the reply direction).
307 * There are two cases to consider:
308 * 1. The packet was DNAT'ed to a device in the same bridge
309 * port group as it was received on. We can still bridge
311 * 2. The packet was DNAT'ed to a different device, either
312 * a non-bridged device or another bridge port group.
313 * The packet will need to be routed.
315 * The correct way of distinguishing between these two cases is to
316 * call ip_route_input() and to look at skb->dst->dev, which is
317 * changed to the destination device if ip_route_input() succeeds.
319 * Let's first consider the case that ip_route_input() succeeds:
321 * If the output device equals the logical bridge device the packet
322 * came in on, we can consider this bridging. The corresponding MAC
323 * address will be obtained in br_nf_pre_routing_finish_bridge.
324 * Otherwise, the packet is considered to be routed and we just
325 * change the destination MAC address so that the packet will
326 * later be passed up to the IP stack to be routed. For a redirected
327 * packet, ip_route_input() will give back the localhost as output device,
328 * which differs from the bridge device.
330 * Let's now consider the case that ip_route_input() fails:
332 * This can be because the destination address is martian, in which case
333 * the packet will be dropped.
334 * If IP forwarding is disabled, ip_route_input() will fail, while
335 * ip_route_output_key() can return success. The source
336 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
337 * thinks we're handling a locally generated packet and won't care
338 * if IP forwarding is enabled. If the output device equals the logical bridge
339 * device, we proceed as if ip_route_input() succeeded. If it differs from the
340 * logical bridge port or if ip_route_output_key() fails we drop the packet.
342 static int br_nf_pre_routing_finish(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
344 struct net_device
*dev
= skb
->dev
;
345 struct iphdr
*iph
= ip_hdr(skb
);
346 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
350 nf_bridge
->frag_max_size
= IPCB(skb
)->frag_max_size
;
352 if (nf_bridge
->pkt_otherhost
) {
353 skb
->pkt_type
= PACKET_OTHERHOST
;
354 nf_bridge
->pkt_otherhost
= false;
356 nf_bridge
->in_prerouting
= 0;
357 if (br_nf_ipv4_daddr_was_changed(skb
, nf_bridge
)) {
358 if ((err
= ip_route_input(skb
, iph
->daddr
, iph
->saddr
, iph
->tos
, dev
))) {
359 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
361 /* If err equals -EHOSTUNREACH the error is due to a
362 * martian destination or due to the fact that
363 * forwarding is disabled. For most martian packets,
364 * ip_route_output_key() will fail. It won't fail for 2 types of
365 * martian destinations: loopback destinations and destination
366 * 0.0.0.0. In both cases the packet will be dropped because the
367 * destination is the loopback device and not the bridge. */
368 if (err
!= -EHOSTUNREACH
|| !in_dev
|| IN_DEV_FORWARD(in_dev
))
371 rt
= ip_route_output(net
, iph
->daddr
, 0,
372 RT_TOS(iph
->tos
), 0);
374 /* - Bridged-and-DNAT'ed traffic doesn't
375 * require ip_forwarding. */
376 if (rt
->dst
.dev
== dev
) {
377 skb_dst_set(skb
, &rt
->dst
);
386 if (skb_dst(skb
)->dev
== dev
) {
388 skb
->dev
= nf_bridge
->physindev
;
389 nf_bridge_update_protocol(skb
);
390 nf_bridge_push_encap_header(skb
);
391 NF_HOOK_THRESH(NFPROTO_BRIDGE
,
393 net
, sk
, skb
, skb
->dev
, NULL
,
394 br_nf_pre_routing_finish_bridge
,
398 ether_addr_copy(eth_hdr(skb
)->h_dest
, dev
->dev_addr
);
399 skb
->pkt_type
= PACKET_HOST
;
402 rt
= bridge_parent_rtable(nf_bridge
->physindev
);
407 skb_dst_set_noref(skb
, &rt
->dst
);
410 skb
->dev
= nf_bridge
->physindev
;
411 nf_bridge_update_protocol(skb
);
412 nf_bridge_push_encap_header(skb
);
413 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_PRE_ROUTING
, net
, sk
, skb
,
415 br_handle_frame_finish
, 1);
420 static struct net_device
*brnf_get_logical_dev(struct sk_buff
*skb
, const struct net_device
*dev
)
422 struct net_device
*vlan
, *br
;
424 br
= bridge_parent(dev
);
425 if (brnf_pass_vlan_indev
== 0 || !skb_vlan_tag_present(skb
))
428 vlan
= __vlan_find_dev_deep_rcu(br
, skb
->vlan_proto
,
429 skb_vlan_tag_get(skb
) & VLAN_VID_MASK
);
431 return vlan
? vlan
: br
;
434 /* Some common code for IPv4/IPv6 */
435 struct net_device
*setup_pre_routing(struct sk_buff
*skb
)
437 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
439 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
440 skb
->pkt_type
= PACKET_HOST
;
441 nf_bridge
->pkt_otherhost
= true;
444 nf_bridge
->in_prerouting
= 1;
445 nf_bridge
->physindev
= skb
->dev
;
446 skb
->dev
= brnf_get_logical_dev(skb
, skb
->dev
);
448 if (skb
->protocol
== htons(ETH_P_8021Q
))
449 nf_bridge
->orig_proto
= BRNF_PROTO_8021Q
;
450 else if (skb
->protocol
== htons(ETH_P_PPP_SES
))
451 nf_bridge
->orig_proto
= BRNF_PROTO_PPPOE
;
453 /* Must drop socket now because of tproxy. */
458 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
459 * Replicate the checks that IPv4 does on packet reception.
460 * Set skb->dev to the bridge device (i.e. parent of the
461 * receiving device) to make netfilter happy, the REDIRECT
462 * target in particular. Save the original destination IP
463 * address to be able to detect DNAT afterwards. */
464 static unsigned int br_nf_pre_routing(void *priv
,
466 const struct nf_hook_state
*state
)
468 struct nf_bridge_info
*nf_bridge
;
469 struct net_bridge_port
*p
;
470 struct net_bridge
*br
;
471 __u32 len
= nf_bridge_encap_header_len(skb
);
473 if (unlikely(!pskb_may_pull(skb
, len
)))
476 p
= br_port_get_rcu(state
->in
);
481 if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
)) {
482 if (!brnf_call_ip6tables
&& !br
->nf_call_ip6tables
)
485 nf_bridge_pull_encap_header_rcsum(skb
);
486 return br_nf_pre_routing_ipv6(priv
, skb
, state
);
489 if (!brnf_call_iptables
&& !br
->nf_call_iptables
)
492 if (!IS_IP(skb
) && !IS_VLAN_IP(skb
) && !IS_PPPOE_IP(skb
))
495 nf_bridge_pull_encap_header_rcsum(skb
);
497 if (br_validate_ipv4(state
->net
, skb
))
500 nf_bridge_put(skb
->nf_bridge
);
501 if (!nf_bridge_alloc(skb
))
503 if (!setup_pre_routing(skb
))
506 nf_bridge
= nf_bridge_info_get(skb
);
507 nf_bridge
->ipv4_daddr
= ip_hdr(skb
)->daddr
;
509 skb
->protocol
= htons(ETH_P_IP
);
511 NF_HOOK(NFPROTO_IPV4
, NF_INET_PRE_ROUTING
, state
->net
, state
->sk
, skb
,
513 br_nf_pre_routing_finish
);
519 /* PF_BRIDGE/LOCAL_IN ************************************************/
520 /* The packet is locally destined, which requires a real
521 * dst_entry, so detach the fake one. On the way up, the
522 * packet would pass through PRE_ROUTING again (which already
523 * took place when the packet entered the bridge), but we
524 * register an IPv4 PRE_ROUTING 'sabotage' hook that will
525 * prevent this from happening. */
526 static unsigned int br_nf_local_in(void *priv
,
528 const struct nf_hook_state
*state
)
530 br_drop_fake_rtable(skb
);
534 /* PF_BRIDGE/FORWARD *************************************************/
535 static int br_nf_forward_finish(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
537 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
538 struct net_device
*in
;
540 if (!IS_ARP(skb
) && !IS_VLAN_ARP(skb
)) {
542 if (skb
->protocol
== htons(ETH_P_IP
))
543 nf_bridge
->frag_max_size
= IPCB(skb
)->frag_max_size
;
545 if (skb
->protocol
== htons(ETH_P_IPV6
))
546 nf_bridge
->frag_max_size
= IP6CB(skb
)->frag_max_size
;
548 in
= nf_bridge
->physindev
;
549 if (nf_bridge
->pkt_otherhost
) {
550 skb
->pkt_type
= PACKET_OTHERHOST
;
551 nf_bridge
->pkt_otherhost
= false;
553 nf_bridge_update_protocol(skb
);
555 in
= *((struct net_device
**)(skb
->cb
));
557 nf_bridge_push_encap_header(skb
);
559 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_FORWARD
, net
, sk
, skb
,
560 in
, skb
->dev
, br_forward_finish
, 1);
565 /* This is the 'purely bridged' case. For IP, we pass the packet to
566 * netfilter with indev and outdev set to the bridge device,
567 * but we are still able to filter on the 'real' indev/outdev
568 * because of the physdev module. For ARP, indev and outdev are the
570 static unsigned int br_nf_forward_ip(void *priv
,
572 const struct nf_hook_state
*state
)
574 struct nf_bridge_info
*nf_bridge
;
575 struct net_device
*parent
;
581 /* Need exclusive nf_bridge_info since we might have multiple
582 * different physoutdevs. */
583 if (!nf_bridge_unshare(skb
))
586 nf_bridge
= nf_bridge_info_get(skb
);
590 parent
= bridge_parent(state
->out
);
594 if (IS_IP(skb
) || IS_VLAN_IP(skb
) || IS_PPPOE_IP(skb
))
596 else if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
))
601 nf_bridge_pull_encap_header(skb
);
603 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
604 skb
->pkt_type
= PACKET_HOST
;
605 nf_bridge
->pkt_otherhost
= true;
608 if (pf
== NFPROTO_IPV4
) {
609 if (br_validate_ipv4(state
->net
, skb
))
611 IPCB(skb
)->frag_max_size
= nf_bridge
->frag_max_size
;
614 if (pf
== NFPROTO_IPV6
) {
615 if (br_validate_ipv6(state
->net
, skb
))
617 IP6CB(skb
)->frag_max_size
= nf_bridge
->frag_max_size
;
620 nf_bridge
->physoutdev
= skb
->dev
;
621 if (pf
== NFPROTO_IPV4
)
622 skb
->protocol
= htons(ETH_P_IP
);
624 skb
->protocol
= htons(ETH_P_IPV6
);
626 NF_HOOK(pf
, NF_INET_FORWARD
, state
->net
, NULL
, skb
,
627 brnf_get_logical_dev(skb
, state
->in
),
628 parent
, br_nf_forward_finish
);
633 static unsigned int br_nf_forward_arp(void *priv
,
635 const struct nf_hook_state
*state
)
637 struct net_bridge_port
*p
;
638 struct net_bridge
*br
;
639 struct net_device
**d
= (struct net_device
**)(skb
->cb
);
641 p
= br_port_get_rcu(state
->out
);
646 if (!brnf_call_arptables
&& !br
->nf_call_arptables
)
650 if (!IS_VLAN_ARP(skb
))
652 nf_bridge_pull_encap_header(skb
);
655 if (arp_hdr(skb
)->ar_pln
!= 4) {
656 if (IS_VLAN_ARP(skb
))
657 nf_bridge_push_encap_header(skb
);
661 NF_HOOK(NFPROTO_ARP
, NF_ARP_FORWARD
, state
->net
, state
->sk
, skb
,
662 state
->in
, state
->out
, br_nf_forward_finish
);
667 static int br_nf_push_frag_xmit(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
669 struct brnf_frag_data
*data
;
672 data
= this_cpu_ptr(&brnf_frag_data_storage
);
673 err
= skb_cow_head(skb
, data
->size
);
680 if (data
->vlan_tci
) {
681 skb
->vlan_tci
= data
->vlan_tci
;
682 skb
->vlan_proto
= data
->vlan_proto
;
685 skb_copy_to_linear_data_offset(skb
, -data
->size
, data
->mac
, data
->size
);
686 __skb_push(skb
, data
->encap_size
);
688 nf_bridge_info_free(skb
);
689 return br_dev_queue_push_xmit(net
, sk
, skb
);
693 br_nf_ip_fragment(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
,
694 int (*output
)(struct net
*, struct sock
*, struct sk_buff
*))
696 unsigned int mtu
= ip_skb_dst_mtu(skb
);
697 struct iphdr
*iph
= ip_hdr(skb
);
699 if (unlikely(((iph
->frag_off
& htons(IP_DF
)) && !skb
->ignore_df
) ||
700 (IPCB(skb
)->frag_max_size
&&
701 IPCB(skb
)->frag_max_size
> mtu
))) {
702 IP_INC_STATS(net
, IPSTATS_MIB_FRAGFAILS
);
707 return ip_do_fragment(net
, sk
, skb
, output
);
710 static unsigned int nf_bridge_mtu_reduction(const struct sk_buff
*skb
)
712 if (skb
->nf_bridge
->orig_proto
== BRNF_PROTO_PPPOE
)
713 return PPPOE_SES_HLEN
;
717 static int br_nf_dev_queue_xmit(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
719 struct nf_bridge_info
*nf_bridge
;
720 unsigned int mtu_reserved
;
722 mtu_reserved
= nf_bridge_mtu_reduction(skb
);
724 if (skb_is_gso(skb
) || skb
->len
+ mtu_reserved
<= skb
->dev
->mtu
) {
725 nf_bridge_info_free(skb
);
726 return br_dev_queue_push_xmit(net
, sk
, skb
);
729 nf_bridge
= nf_bridge_info_get(skb
);
731 /* This is wrong! We should preserve the original fragment
732 * boundaries by preserving frag_list rather than refragmenting.
734 if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV4
) &&
735 skb
->protocol
== htons(ETH_P_IP
)) {
736 struct brnf_frag_data
*data
;
738 if (br_validate_ipv4(net
, skb
))
741 IPCB(skb
)->frag_max_size
= nf_bridge
->frag_max_size
;
743 nf_bridge_update_protocol(skb
);
745 data
= this_cpu_ptr(&brnf_frag_data_storage
);
747 data
->vlan_tci
= skb
->vlan_tci
;
748 data
->vlan_proto
= skb
->vlan_proto
;
749 data
->encap_size
= nf_bridge_encap_header_len(skb
);
750 data
->size
= ETH_HLEN
+ data
->encap_size
;
752 skb_copy_from_linear_data_offset(skb
, -data
->size
, data
->mac
,
755 return br_nf_ip_fragment(net
, sk
, skb
, br_nf_push_frag_xmit
);
757 if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV6
) &&
758 skb
->protocol
== htons(ETH_P_IPV6
)) {
759 const struct nf_ipv6_ops
*v6ops
= nf_get_ipv6_ops();
760 struct brnf_frag_data
*data
;
762 if (br_validate_ipv6(net
, skb
))
765 IP6CB(skb
)->frag_max_size
= nf_bridge
->frag_max_size
;
767 nf_bridge_update_protocol(skb
);
769 data
= this_cpu_ptr(&brnf_frag_data_storage
);
770 data
->encap_size
= nf_bridge_encap_header_len(skb
);
771 data
->size
= ETH_HLEN
+ data
->encap_size
;
773 skb_copy_from_linear_data_offset(skb
, -data
->size
, data
->mac
,
777 return v6ops
->fragment(net
, sk
, skb
, br_nf_push_frag_xmit
);
782 nf_bridge_info_free(skb
);
783 return br_dev_queue_push_xmit(net
, sk
, skb
);
789 /* PF_BRIDGE/POST_ROUTING ********************************************/
790 static unsigned int br_nf_post_routing(void *priv
,
792 const struct nf_hook_state
*state
)
794 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
795 struct net_device
*realoutdev
= bridge_parent(skb
->dev
);
798 /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
799 * on a bridge, but was delivered locally and is now being routed:
801 * POST_ROUTING was already invoked from the ip stack.
803 if (!nf_bridge
|| !nf_bridge
->physoutdev
)
809 if (IS_IP(skb
) || IS_VLAN_IP(skb
) || IS_PPPOE_IP(skb
))
811 else if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
))
816 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
817 * about the value of skb->pkt_type. */
818 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
819 skb
->pkt_type
= PACKET_HOST
;
820 nf_bridge
->pkt_otherhost
= true;
823 nf_bridge_pull_encap_header(skb
);
824 if (pf
== NFPROTO_IPV4
)
825 skb
->protocol
= htons(ETH_P_IP
);
827 skb
->protocol
= htons(ETH_P_IPV6
);
829 NF_HOOK(pf
, NF_INET_POST_ROUTING
, state
->net
, state
->sk
, skb
,
831 br_nf_dev_queue_xmit
);
836 /* IP/SABOTAGE *****************************************************/
837 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
838 * for the second time. */
839 static unsigned int ip_sabotage_in(void *priv
,
841 const struct nf_hook_state
*state
)
843 if (skb
->nf_bridge
&& !skb
->nf_bridge
->in_prerouting
)
849 /* This is called when br_netfilter has called into iptables/netfilter,
850 * and DNAT has taken place on a bridge-forwarded packet.
852 * neigh->output has created a new MAC header, with local br0 MAC
855 * This restores the original MAC saddr of the bridged packet
856 * before invoking bridge forward logic to transmit the packet.
858 static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff
*skb
)
860 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
862 skb_pull(skb
, ETH_HLEN
);
863 nf_bridge
->bridged_dnat
= 0;
865 BUILD_BUG_ON(sizeof(nf_bridge
->neigh_header
) != (ETH_HLEN
- ETH_ALEN
));
867 skb_copy_to_linear_data_offset(skb
, -(ETH_HLEN
- ETH_ALEN
),
868 nf_bridge
->neigh_header
,
869 ETH_HLEN
- ETH_ALEN
);
870 skb
->dev
= nf_bridge
->physindev
;
872 nf_bridge
->physoutdev
= NULL
;
873 br_handle_frame_finish(dev_net(skb
->dev
), NULL
, skb
);
876 static int br_nf_dev_xmit(struct sk_buff
*skb
)
878 if (skb
->nf_bridge
&& skb
->nf_bridge
->bridged_dnat
) {
879 br_nf_pre_routing_finish_bridge_slow(skb
);
885 static const struct nf_br_ops br_ops
= {
886 .br_dev_xmit_hook
= br_nf_dev_xmit
,
889 void br_netfilter_enable(void)
892 EXPORT_SYMBOL_GPL(br_netfilter_enable
);
894 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
895 * br_dev_queue_push_xmit is called afterwards */
896 static struct nf_hook_ops br_nf_ops
[] __read_mostly
= {
898 .hook
= br_nf_pre_routing
,
899 .pf
= NFPROTO_BRIDGE
,
900 .hooknum
= NF_BR_PRE_ROUTING
,
901 .priority
= NF_BR_PRI_BRNF
,
904 .hook
= br_nf_local_in
,
905 .pf
= NFPROTO_BRIDGE
,
906 .hooknum
= NF_BR_LOCAL_IN
,
907 .priority
= NF_BR_PRI_BRNF
,
910 .hook
= br_nf_forward_ip
,
911 .pf
= NFPROTO_BRIDGE
,
912 .hooknum
= NF_BR_FORWARD
,
913 .priority
= NF_BR_PRI_BRNF
- 1,
916 .hook
= br_nf_forward_arp
,
917 .pf
= NFPROTO_BRIDGE
,
918 .hooknum
= NF_BR_FORWARD
,
919 .priority
= NF_BR_PRI_BRNF
,
922 .hook
= br_nf_post_routing
,
923 .pf
= NFPROTO_BRIDGE
,
924 .hooknum
= NF_BR_POST_ROUTING
,
925 .priority
= NF_BR_PRI_LAST
,
928 .hook
= ip_sabotage_in
,
930 .hooknum
= NF_INET_PRE_ROUTING
,
931 .priority
= NF_IP_PRI_FIRST
,
934 .hook
= ip_sabotage_in
,
936 .hooknum
= NF_INET_PRE_ROUTING
,
937 .priority
= NF_IP6_PRI_FIRST
,
943 int brnf_sysctl_call_tables(struct ctl_table
*ctl
, int write
,
944 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
948 ret
= proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
950 if (write
&& *(int *)(ctl
->data
))
951 *(int *)(ctl
->data
) = 1;
955 static struct ctl_table brnf_table
[] = {
957 .procname
= "bridge-nf-call-arptables",
958 .data
= &brnf_call_arptables
,
959 .maxlen
= sizeof(int),
961 .proc_handler
= brnf_sysctl_call_tables
,
964 .procname
= "bridge-nf-call-iptables",
965 .data
= &brnf_call_iptables
,
966 .maxlen
= sizeof(int),
968 .proc_handler
= brnf_sysctl_call_tables
,
971 .procname
= "bridge-nf-call-ip6tables",
972 .data
= &brnf_call_ip6tables
,
973 .maxlen
= sizeof(int),
975 .proc_handler
= brnf_sysctl_call_tables
,
978 .procname
= "bridge-nf-filter-vlan-tagged",
979 .data
= &brnf_filter_vlan_tagged
,
980 .maxlen
= sizeof(int),
982 .proc_handler
= brnf_sysctl_call_tables
,
985 .procname
= "bridge-nf-filter-pppoe-tagged",
986 .data
= &brnf_filter_pppoe_tagged
,
987 .maxlen
= sizeof(int),
989 .proc_handler
= brnf_sysctl_call_tables
,
992 .procname
= "bridge-nf-pass-vlan-input-dev",
993 .data
= &brnf_pass_vlan_indev
,
994 .maxlen
= sizeof(int),
996 .proc_handler
= brnf_sysctl_call_tables
,
1002 static int __init
br_netfilter_init(void)
1006 ret
= nf_register_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
1010 #ifdef CONFIG_SYSCTL
1011 brnf_sysctl_header
= register_net_sysctl(&init_net
, "net/bridge", brnf_table
);
1012 if (brnf_sysctl_header
== NULL
) {
1014 "br_netfilter: can't register to sysctl.\n");
1015 nf_unregister_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
1019 RCU_INIT_POINTER(nf_br_ops
, &br_ops
);
1020 printk(KERN_NOTICE
"Bridge firewalling registered\n");
1024 static void __exit
br_netfilter_fini(void)
1026 RCU_INIT_POINTER(nf_br_ops
, NULL
);
1027 nf_unregister_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
1028 #ifdef CONFIG_SYSCTL
1029 unregister_net_sysctl_table(brnf_sysctl_header
);
1033 module_init(br_netfilter_init
);
1034 module_exit(br_netfilter_fini
);
1036 MODULE_LICENSE("GPL");
1037 MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
1038 MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
1039 MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");