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 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
41 #include <net/netfilter/nf_conntrack.h>
44 #include <asm/uaccess.h>
45 #include "br_private.h"
47 #include <linux/sysctl.h>
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;
57 static int brnf_pass_vlan_indev __read_mostly
= 0;
59 #define brnf_call_iptables 1
60 #define brnf_call_ip6tables 1
61 #define brnf_call_arptables 1
62 #define brnf_filter_vlan_tagged 0
63 #define brnf_filter_pppoe_tagged 0
64 #define brnf_pass_vlan_indev 0
68 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP))
70 #define IS_IPV6(skb) \
71 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6))
74 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP))
76 static inline __be16
vlan_proto(const struct sk_buff
*skb
)
78 if (skb_vlan_tag_present(skb
))
80 else if (skb
->protocol
== htons(ETH_P_8021Q
))
81 return vlan_eth_hdr(skb
)->h_vlan_encapsulated_proto
;
86 #define IS_VLAN_IP(skb) \
87 (vlan_proto(skb) == htons(ETH_P_IP) && \
88 brnf_filter_vlan_tagged)
90 #define IS_VLAN_IPV6(skb) \
91 (vlan_proto(skb) == htons(ETH_P_IPV6) && \
92 brnf_filter_vlan_tagged)
94 #define IS_VLAN_ARP(skb) \
95 (vlan_proto(skb) == htons(ETH_P_ARP) && \
96 brnf_filter_vlan_tagged)
98 static inline __be16
pppoe_proto(const struct sk_buff
*skb
)
100 return *((__be16
*)(skb_mac_header(skb
) + ETH_HLEN
+
101 sizeof(struct pppoe_hdr
)));
104 #define IS_PPPOE_IP(skb) \
105 (skb->protocol == htons(ETH_P_PPP_SES) && \
106 pppoe_proto(skb) == htons(PPP_IP) && \
107 brnf_filter_pppoe_tagged)
109 #define IS_PPPOE_IPV6(skb) \
110 (skb->protocol == htons(ETH_P_PPP_SES) && \
111 pppoe_proto(skb) == htons(PPP_IPV6) && \
112 brnf_filter_pppoe_tagged)
114 /* largest possible L2 header, see br_nf_dev_queue_xmit() */
115 #define NF_BRIDGE_MAX_MAC_HEADER_LENGTH (PPPOE_SES_HLEN + ETH_HLEN)
117 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4)
118 struct brnf_frag_data
{
119 char mac
[NF_BRIDGE_MAX_MAC_HEADER_LENGTH
];
124 static DEFINE_PER_CPU(struct brnf_frag_data
, brnf_frag_data_storage
);
127 static struct nf_bridge_info
*nf_bridge_info_get(const struct sk_buff
*skb
)
129 return skb
->nf_bridge
;
132 static inline struct rtable
*bridge_parent_rtable(const struct net_device
*dev
)
134 struct net_bridge_port
*port
;
136 port
= br_port_get_rcu(dev
);
137 return port
? &port
->br
->fake_rtable
: NULL
;
140 static inline struct net_device
*bridge_parent(const struct net_device
*dev
)
142 struct net_bridge_port
*port
;
144 port
= br_port_get_rcu(dev
);
145 return port
? port
->br
->dev
: NULL
;
148 static inline struct nf_bridge_info
*nf_bridge_alloc(struct sk_buff
*skb
)
150 skb
->nf_bridge
= kzalloc(sizeof(struct nf_bridge_info
), GFP_ATOMIC
);
151 if (likely(skb
->nf_bridge
))
152 atomic_set(&(skb
->nf_bridge
->use
), 1);
154 return skb
->nf_bridge
;
157 static inline struct nf_bridge_info
*nf_bridge_unshare(struct sk_buff
*skb
)
159 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
161 if (atomic_read(&nf_bridge
->use
) > 1) {
162 struct nf_bridge_info
*tmp
= nf_bridge_alloc(skb
);
165 memcpy(tmp
, nf_bridge
, sizeof(struct nf_bridge_info
));
166 atomic_set(&tmp
->use
, 1);
168 nf_bridge_put(nf_bridge
);
174 static unsigned int nf_bridge_encap_header_len(const struct sk_buff
*skb
)
176 switch (skb
->protocol
) {
177 case __cpu_to_be16(ETH_P_8021Q
):
179 case __cpu_to_be16(ETH_P_PPP_SES
):
180 return PPPOE_SES_HLEN
;
186 static inline void nf_bridge_push_encap_header(struct sk_buff
*skb
)
188 unsigned int len
= nf_bridge_encap_header_len(skb
);
191 skb
->network_header
-= len
;
194 static inline void nf_bridge_pull_encap_header(struct sk_buff
*skb
)
196 unsigned int len
= nf_bridge_encap_header_len(skb
);
199 skb
->network_header
+= len
;
202 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff
*skb
)
204 unsigned int len
= nf_bridge_encap_header_len(skb
);
206 skb_pull_rcsum(skb
, len
);
207 skb
->network_header
+= len
;
210 /* When handing a packet over to the IP layer
211 * check whether we have a skb that is in the
215 static int br_parse_ip_options(struct sk_buff
*skb
)
217 const struct iphdr
*iph
;
218 struct net_device
*dev
= skb
->dev
;
221 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
226 /* Basic sanity checks */
227 if (iph
->ihl
< 5 || iph
->version
!= 4)
230 if (!pskb_may_pull(skb
, iph
->ihl
*4))
234 if (unlikely(ip_fast_csum((u8
*)iph
, iph
->ihl
)))
237 len
= ntohs(iph
->tot_len
);
238 if (skb
->len
< len
) {
239 IP_INC_STATS_BH(dev_net(dev
), IPSTATS_MIB_INTRUNCATEDPKTS
);
241 } else if (len
< (iph
->ihl
*4))
244 if (pskb_trim_rcsum(skb
, len
)) {
245 IP_INC_STATS_BH(dev_net(dev
), IPSTATS_MIB_INDISCARDS
);
249 memset(IPCB(skb
), 0, sizeof(struct inet_skb_parm
));
250 /* We should really parse IP options here but until
251 * somebody who actually uses IP options complains to
252 * us we'll just silently ignore the options because
258 IP_INC_STATS_BH(dev_net(dev
), IPSTATS_MIB_INHDRERRORS
);
263 static void nf_bridge_update_protocol(struct sk_buff
*skb
)
265 if (skb
->nf_bridge
->mask
& BRNF_8021Q
)
266 skb
->protocol
= htons(ETH_P_8021Q
);
267 else if (skb
->nf_bridge
->mask
& BRNF_PPPoE
)
268 skb
->protocol
= htons(ETH_P_PPP_SES
);
271 /* PF_BRIDGE/PRE_ROUTING *********************************************/
272 /* Undo the changes made for ip6tables PREROUTING and continue the
273 * bridge PRE_ROUTING hook. */
274 static int br_nf_pre_routing_finish_ipv6(struct sk_buff
*skb
)
276 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
279 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
280 skb
->pkt_type
= PACKET_OTHERHOST
;
281 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
283 nf_bridge
->mask
^= BRNF_NF_BRIDGE_PREROUTING
;
285 rt
= bridge_parent_rtable(nf_bridge
->physindev
);
290 skb_dst_set_noref(skb
, &rt
->dst
);
292 skb
->dev
= nf_bridge
->physindev
;
293 nf_bridge_update_protocol(skb
);
294 nf_bridge_push_encap_header(skb
);
295 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
296 br_handle_frame_finish
, 1);
301 /* Obtain the correct destination MAC address, while preserving the original
302 * source MAC address. If we already know this address, we just copy it. If we
303 * don't, we use the neighbour framework to find out. In both cases, we make
304 * sure that br_handle_frame_finish() is called afterwards.
306 static int br_nf_pre_routing_finish_bridge(struct sk_buff
*skb
)
308 struct neighbour
*neigh
;
309 struct dst_entry
*dst
;
311 skb
->dev
= bridge_parent(skb
->dev
);
315 neigh
= dst_neigh_lookup_skb(dst
, skb
);
317 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
320 if (neigh
->hh
.hh_len
) {
321 neigh_hh_bridge(&neigh
->hh
, skb
);
322 skb
->dev
= nf_bridge
->physindev
;
323 ret
= br_handle_frame_finish(skb
);
325 /* the neighbour function below overwrites the complete
326 * MAC header, so we save the Ethernet source address and
329 skb_copy_from_linear_data_offset(skb
,
330 -(ETH_HLEN
-ETH_ALEN
),
331 nf_bridge
->neigh_header
,
333 /* tell br_dev_xmit to continue with forwarding */
334 nf_bridge
->mask
|= BRNF_BRIDGED_DNAT
;
335 /* FIXME Need to refragment */
336 ret
= neigh
->output(neigh
, skb
);
338 neigh_release(neigh
);
346 static bool dnat_took_place(const struct sk_buff
*skb
)
348 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
349 enum ip_conntrack_info ctinfo
;
352 ct
= nf_ct_get(skb
, &ctinfo
);
353 if (!ct
|| nf_ct_is_untracked(ct
))
356 return test_bit(IPS_DST_NAT_BIT
, &ct
->status
);
362 /* This requires some explaining. If DNAT has taken place,
363 * we will need to fix up the destination Ethernet address.
365 * There are two cases to consider:
366 * 1. The packet was DNAT'ed to a device in the same bridge
367 * port group as it was received on. We can still bridge
369 * 2. The packet was DNAT'ed to a different device, either
370 * a non-bridged device or another bridge port group.
371 * The packet will need to be routed.
373 * The correct way of distinguishing between these two cases is to
374 * call ip_route_input() and to look at skb->dst->dev, which is
375 * changed to the destination device if ip_route_input() succeeds.
377 * Let's first consider the case that ip_route_input() succeeds:
379 * If the output device equals the logical bridge device the packet
380 * came in on, we can consider this bridging. The corresponding MAC
381 * address will be obtained in br_nf_pre_routing_finish_bridge.
382 * Otherwise, the packet is considered to be routed and we just
383 * change the destination MAC address so that the packet will
384 * later be passed up to the IP stack to be routed. For a redirected
385 * packet, ip_route_input() will give back the localhost as output device,
386 * which differs from the bridge device.
388 * Let's now consider the case that ip_route_input() fails:
390 * This can be because the destination address is martian, in which case
391 * the packet will be dropped.
392 * If IP forwarding is disabled, ip_route_input() will fail, while
393 * ip_route_output_key() can return success. The source
394 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
395 * thinks we're handling a locally generated packet and won't care
396 * if IP forwarding is enabled. If the output device equals the logical bridge
397 * device, we proceed as if ip_route_input() succeeded. If it differs from the
398 * logical bridge port or if ip_route_output_key() fails we drop the packet.
400 static int br_nf_pre_routing_finish(struct sk_buff
*skb
)
402 struct net_device
*dev
= skb
->dev
;
403 struct iphdr
*iph
= ip_hdr(skb
);
404 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
409 frag_max_size
= IPCB(skb
)->frag_max_size
;
410 BR_INPUT_SKB_CB(skb
)->frag_max_size
= frag_max_size
;
412 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
413 skb
->pkt_type
= PACKET_OTHERHOST
;
414 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
416 nf_bridge
->mask
^= BRNF_NF_BRIDGE_PREROUTING
;
417 if (dnat_took_place(skb
)) {
418 if ((err
= ip_route_input(skb
, iph
->daddr
, iph
->saddr
, iph
->tos
, dev
))) {
419 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
421 /* If err equals -EHOSTUNREACH the error is due to a
422 * martian destination or due to the fact that
423 * forwarding is disabled. For most martian packets,
424 * ip_route_output_key() will fail. It won't fail for 2 types of
425 * martian destinations: loopback destinations and destination
426 * 0.0.0.0. In both cases the packet will be dropped because the
427 * destination is the loopback device and not the bridge. */
428 if (err
!= -EHOSTUNREACH
|| !in_dev
|| IN_DEV_FORWARD(in_dev
))
431 rt
= ip_route_output(dev_net(dev
), iph
->daddr
, 0,
432 RT_TOS(iph
->tos
), 0);
434 /* - Bridged-and-DNAT'ed traffic doesn't
435 * require ip_forwarding. */
436 if (rt
->dst
.dev
== dev
) {
437 skb_dst_set(skb
, &rt
->dst
);
446 if (skb_dst(skb
)->dev
== dev
) {
448 skb
->dev
= nf_bridge
->physindev
;
449 nf_bridge_update_protocol(skb
);
450 nf_bridge_push_encap_header(skb
);
451 NF_HOOK_THRESH(NFPROTO_BRIDGE
,
454 br_nf_pre_routing_finish_bridge
,
458 ether_addr_copy(eth_hdr(skb
)->h_dest
, dev
->dev_addr
);
459 skb
->pkt_type
= PACKET_HOST
;
462 rt
= bridge_parent_rtable(nf_bridge
->physindev
);
467 skb_dst_set_noref(skb
, &rt
->dst
);
470 skb
->dev
= nf_bridge
->physindev
;
471 nf_bridge_update_protocol(skb
);
472 nf_bridge_push_encap_header(skb
);
473 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
474 br_handle_frame_finish
, 1);
479 static struct net_device
*brnf_get_logical_dev(struct sk_buff
*skb
, const struct net_device
*dev
)
481 struct net_device
*vlan
, *br
;
483 br
= bridge_parent(dev
);
484 if (brnf_pass_vlan_indev
== 0 || !skb_vlan_tag_present(skb
))
487 vlan
= __vlan_find_dev_deep_rcu(br
, skb
->vlan_proto
,
488 skb_vlan_tag_get(skb
) & VLAN_VID_MASK
);
490 return vlan
? vlan
: br
;
493 /* Some common code for IPv4/IPv6 */
494 static struct net_device
*setup_pre_routing(struct sk_buff
*skb
)
496 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
498 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
499 skb
->pkt_type
= PACKET_HOST
;
500 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
503 nf_bridge
->mask
|= BRNF_NF_BRIDGE_PREROUTING
;
504 nf_bridge
->physindev
= skb
->dev
;
505 skb
->dev
= brnf_get_logical_dev(skb
, skb
->dev
);
506 if (skb
->protocol
== htons(ETH_P_8021Q
))
507 nf_bridge
->mask
|= BRNF_8021Q
;
508 else if (skb
->protocol
== htons(ETH_P_PPP_SES
))
509 nf_bridge
->mask
|= BRNF_PPPoE
;
511 /* Must drop socket now because of tproxy. */
516 /* We only check the length. A bridge shouldn't do any hop-by-hop stuff anyway */
517 static int check_hbh_len(struct sk_buff
*skb
)
519 unsigned char *raw
= (u8
*)(ipv6_hdr(skb
) + 1);
521 const unsigned char *nh
= skb_network_header(skb
);
523 int len
= (raw
[1] + 1) << 3;
525 if ((raw
+ len
) - skb
->data
> skb_headlen(skb
))
532 int optlen
= nh
[off
+ 1] + 2;
543 if (nh
[off
+ 1] != 4 || (off
& 3) != 2)
545 pkt_len
= ntohl(*(__be32
*) (nh
+ off
+ 2));
546 if (pkt_len
<= IPV6_MAXPLEN
||
547 ipv6_hdr(skb
)->payload_len
)
549 if (pkt_len
> skb
->len
- sizeof(struct ipv6hdr
))
551 if (pskb_trim_rcsum(skb
,
552 pkt_len
+ sizeof(struct ipv6hdr
)))
554 nh
= skb_network_header(skb
);
571 /* Replicate the checks that IPv6 does on packet reception and pass the packet
572 * to ip6tables, which doesn't support NAT, so things are fairly simple. */
573 static unsigned int br_nf_pre_routing_ipv6(const struct nf_hook_ops
*ops
,
575 const struct net_device
*in
,
576 const struct net_device
*out
,
577 int (*okfn
)(struct sk_buff
*))
579 const struct ipv6hdr
*hdr
;
582 if (skb
->len
< sizeof(struct ipv6hdr
))
585 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
590 if (hdr
->version
!= 6)
593 pkt_len
= ntohs(hdr
->payload_len
);
595 if (pkt_len
|| hdr
->nexthdr
!= NEXTHDR_HOP
) {
596 if (pkt_len
+ sizeof(struct ipv6hdr
) > skb
->len
)
598 if (pskb_trim_rcsum(skb
, pkt_len
+ sizeof(struct ipv6hdr
)))
601 if (hdr
->nexthdr
== NEXTHDR_HOP
&& check_hbh_len(skb
))
604 nf_bridge_put(skb
->nf_bridge
);
605 if (!nf_bridge_alloc(skb
))
607 if (!setup_pre_routing(skb
))
610 skb
->protocol
= htons(ETH_P_IPV6
);
611 NF_HOOK(NFPROTO_IPV6
, NF_INET_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
612 br_nf_pre_routing_finish_ipv6
);
617 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
618 * Replicate the checks that IPv4 does on packet reception.
619 * Set skb->dev to the bridge device (i.e. parent of the
620 * receiving device) to make netfilter happy, the REDIRECT
621 * target in particular. Save the original destination IP
622 * address to be able to detect DNAT afterwards. */
623 static unsigned int br_nf_pre_routing(const struct nf_hook_ops
*ops
,
625 const struct net_device
*in
,
626 const struct net_device
*out
,
627 int (*okfn
)(struct sk_buff
*))
629 struct net_bridge_port
*p
;
630 struct net_bridge
*br
;
631 __u32 len
= nf_bridge_encap_header_len(skb
);
633 if (unlikely(!pskb_may_pull(skb
, len
)))
636 p
= br_port_get_rcu(in
);
641 if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
)) {
642 if (!brnf_call_ip6tables
&& !br
->nf_call_ip6tables
)
645 nf_bridge_pull_encap_header_rcsum(skb
);
646 return br_nf_pre_routing_ipv6(ops
, skb
, in
, out
, okfn
);
649 if (!brnf_call_iptables
&& !br
->nf_call_iptables
)
652 if (!IS_IP(skb
) && !IS_VLAN_IP(skb
) && !IS_PPPOE_IP(skb
))
655 nf_bridge_pull_encap_header_rcsum(skb
);
657 if (br_parse_ip_options(skb
))
660 nf_bridge_put(skb
->nf_bridge
);
661 if (!nf_bridge_alloc(skb
))
663 if (!setup_pre_routing(skb
))
666 skb
->protocol
= htons(ETH_P_IP
);
668 NF_HOOK(NFPROTO_IPV4
, NF_INET_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
669 br_nf_pre_routing_finish
);
675 /* PF_BRIDGE/LOCAL_IN ************************************************/
676 /* The packet is locally destined, which requires a real
677 * dst_entry, so detach the fake one. On the way up, the
678 * packet would pass through PRE_ROUTING again (which already
679 * took place when the packet entered the bridge), but we
680 * register an IPv4 PRE_ROUTING 'sabotage' hook that will
681 * prevent this from happening. */
682 static unsigned int br_nf_local_in(const struct nf_hook_ops
*ops
,
684 const struct net_device
*in
,
685 const struct net_device
*out
,
686 int (*okfn
)(struct sk_buff
*))
688 br_drop_fake_rtable(skb
);
692 /* PF_BRIDGE/FORWARD *************************************************/
693 static int br_nf_forward_finish(struct sk_buff
*skb
)
695 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
696 struct net_device
*in
;
698 if (!IS_ARP(skb
) && !IS_VLAN_ARP(skb
)) {
701 if (skb
->protocol
== htons(ETH_P_IP
)) {
702 frag_max_size
= IPCB(skb
)->frag_max_size
;
703 BR_INPUT_SKB_CB(skb
)->frag_max_size
= frag_max_size
;
706 in
= nf_bridge
->physindev
;
707 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
708 skb
->pkt_type
= PACKET_OTHERHOST
;
709 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
711 nf_bridge_update_protocol(skb
);
713 in
= *((struct net_device
**)(skb
->cb
));
715 nf_bridge_push_encap_header(skb
);
717 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_FORWARD
, skb
, in
,
718 skb
->dev
, br_forward_finish
, 1);
723 /* This is the 'purely bridged' case. For IP, we pass the packet to
724 * netfilter with indev and outdev set to the bridge device,
725 * but we are still able to filter on the 'real' indev/outdev
726 * because of the physdev module. For ARP, indev and outdev are the
728 static unsigned int br_nf_forward_ip(const struct nf_hook_ops
*ops
,
730 const struct net_device
*in
,
731 const struct net_device
*out
,
732 int (*okfn
)(struct sk_buff
*))
734 struct nf_bridge_info
*nf_bridge
;
735 struct net_device
*parent
;
741 /* Need exclusive nf_bridge_info since we might have multiple
742 * different physoutdevs. */
743 if (!nf_bridge_unshare(skb
))
746 nf_bridge
= nf_bridge_info_get(skb
);
750 parent
= bridge_parent(out
);
754 if (IS_IP(skb
) || IS_VLAN_IP(skb
) || IS_PPPOE_IP(skb
))
756 else if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
))
761 nf_bridge_pull_encap_header(skb
);
763 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
764 skb
->pkt_type
= PACKET_HOST
;
765 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
768 if (pf
== NFPROTO_IPV4
) {
769 int frag_max
= BR_INPUT_SKB_CB(skb
)->frag_max_size
;
771 if (br_parse_ip_options(skb
))
774 IPCB(skb
)->frag_max_size
= frag_max
;
777 nf_bridge
->physoutdev
= skb
->dev
;
778 if (pf
== NFPROTO_IPV4
)
779 skb
->protocol
= htons(ETH_P_IP
);
781 skb
->protocol
= htons(ETH_P_IPV6
);
783 NF_HOOK(pf
, NF_INET_FORWARD
, skb
, brnf_get_logical_dev(skb
, in
), parent
,
784 br_nf_forward_finish
);
789 static unsigned int br_nf_forward_arp(const struct nf_hook_ops
*ops
,
791 const struct net_device
*in
,
792 const struct net_device
*out
,
793 int (*okfn
)(struct sk_buff
*))
795 struct net_bridge_port
*p
;
796 struct net_bridge
*br
;
797 struct net_device
**d
= (struct net_device
**)(skb
->cb
);
799 p
= br_port_get_rcu(out
);
804 if (!brnf_call_arptables
&& !br
->nf_call_arptables
)
808 if (!IS_VLAN_ARP(skb
))
810 nf_bridge_pull_encap_header(skb
);
813 if (arp_hdr(skb
)->ar_pln
!= 4) {
814 if (IS_VLAN_ARP(skb
))
815 nf_bridge_push_encap_header(skb
);
818 *d
= (struct net_device
*)in
;
819 NF_HOOK(NFPROTO_ARP
, NF_ARP_FORWARD
, skb
, (struct net_device
*)in
,
820 (struct net_device
*)out
, br_nf_forward_finish
);
825 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4)
826 static int br_nf_push_frag_xmit(struct sk_buff
*skb
)
828 struct brnf_frag_data
*data
;
831 data
= this_cpu_ptr(&brnf_frag_data_storage
);
832 err
= skb_cow_head(skb
, data
->size
);
839 skb_copy_to_linear_data_offset(skb
, -data
->size
, data
->mac
, data
->size
);
840 __skb_push(skb
, data
->encap_size
);
842 return br_dev_queue_push_xmit(skb
);
845 static int br_nf_dev_queue_xmit(struct sk_buff
*skb
)
849 unsigned int mtu_reserved
;
851 if (skb_is_gso(skb
) || skb
->protocol
!= htons(ETH_P_IP
))
852 return br_dev_queue_push_xmit(skb
);
854 mtu_reserved
= nf_bridge_mtu_reduction(skb
);
855 /* This is wrong! We should preserve the original fragment
856 * boundaries by preserving frag_list rather than refragmenting.
858 if (skb
->len
+ mtu_reserved
> skb
->dev
->mtu
) {
859 struct brnf_frag_data
*data
;
861 frag_max_size
= BR_INPUT_SKB_CB(skb
)->frag_max_size
;
862 if (br_parse_ip_options(skb
))
863 /* Drop invalid packet */
865 IPCB(skb
)->frag_max_size
= frag_max_size
;
867 nf_bridge_update_protocol(skb
);
869 data
= this_cpu_ptr(&brnf_frag_data_storage
);
870 data
->encap_size
= nf_bridge_encap_header_len(skb
);
871 data
->size
= ETH_HLEN
+ data
->encap_size
;
873 skb_copy_from_linear_data_offset(skb
, -data
->size
, data
->mac
,
876 ret
= ip_fragment(skb
, br_nf_push_frag_xmit
);
878 ret
= br_dev_queue_push_xmit(skb
);
884 static int br_nf_dev_queue_xmit(struct sk_buff
*skb
)
886 return br_dev_queue_push_xmit(skb
);
890 /* PF_BRIDGE/POST_ROUTING ********************************************/
891 static unsigned int br_nf_post_routing(const struct nf_hook_ops
*ops
,
893 const struct net_device
*in
,
894 const struct net_device
*out
,
895 int (*okfn
)(struct sk_buff
*))
897 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
898 struct net_device
*realoutdev
= bridge_parent(skb
->dev
);
901 /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
902 * on a bridge, but was delivered locally and is now being routed:
904 * POST_ROUTING was already invoked from the ip stack.
906 if (!nf_bridge
|| !nf_bridge
->physoutdev
)
912 if (IS_IP(skb
) || IS_VLAN_IP(skb
) || IS_PPPOE_IP(skb
))
914 else if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
))
919 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
920 * about the value of skb->pkt_type. */
921 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
922 skb
->pkt_type
= PACKET_HOST
;
923 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
926 nf_bridge_pull_encap_header(skb
);
927 if (pf
== NFPROTO_IPV4
)
928 skb
->protocol
= htons(ETH_P_IP
);
930 skb
->protocol
= htons(ETH_P_IPV6
);
932 NF_HOOK(pf
, NF_INET_POST_ROUTING
, skb
, NULL
, realoutdev
,
933 br_nf_dev_queue_xmit
);
938 /* IP/SABOTAGE *****************************************************/
939 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
940 * for the second time. */
941 static unsigned int ip_sabotage_in(const struct nf_hook_ops
*ops
,
943 const struct net_device
*in
,
944 const struct net_device
*out
,
945 int (*okfn
)(struct sk_buff
*))
947 if (skb
->nf_bridge
&&
948 !(skb
->nf_bridge
->mask
& BRNF_NF_BRIDGE_PREROUTING
)) {
955 /* This is called when br_netfilter has called into iptables/netfilter,
956 * and DNAT has taken place on a bridge-forwarded packet.
958 * neigh->output has created a new MAC header, with local br0 MAC
961 * This restores the original MAC saddr of the bridged packet
962 * before invoking bridge forward logic to transmit the packet.
964 static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff
*skb
)
966 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
968 skb_pull(skb
, ETH_HLEN
);
969 nf_bridge
->mask
&= ~BRNF_BRIDGED_DNAT
;
971 BUILD_BUG_ON(sizeof(nf_bridge
->neigh_header
) != (ETH_HLEN
- ETH_ALEN
));
973 skb_copy_to_linear_data_offset(skb
, -(ETH_HLEN
- ETH_ALEN
),
974 nf_bridge
->neigh_header
,
975 ETH_HLEN
- ETH_ALEN
);
976 skb
->dev
= nf_bridge
->physindev
;
977 br_handle_frame_finish(skb
);
980 static int br_nf_dev_xmit(struct sk_buff
*skb
)
982 if (skb
->nf_bridge
&& (skb
->nf_bridge
->mask
& BRNF_BRIDGED_DNAT
)) {
983 br_nf_pre_routing_finish_bridge_slow(skb
);
989 static const struct nf_br_ops br_ops
= {
990 .br_dev_xmit_hook
= br_nf_dev_xmit
,
993 void br_netfilter_enable(void)
996 EXPORT_SYMBOL_GPL(br_netfilter_enable
);
998 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
999 * br_dev_queue_push_xmit is called afterwards */
1000 static struct nf_hook_ops br_nf_ops
[] __read_mostly
= {
1002 .hook
= br_nf_pre_routing
,
1003 .owner
= THIS_MODULE
,
1004 .pf
= NFPROTO_BRIDGE
,
1005 .hooknum
= NF_BR_PRE_ROUTING
,
1006 .priority
= NF_BR_PRI_BRNF
,
1009 .hook
= br_nf_local_in
,
1010 .owner
= THIS_MODULE
,
1011 .pf
= NFPROTO_BRIDGE
,
1012 .hooknum
= NF_BR_LOCAL_IN
,
1013 .priority
= NF_BR_PRI_BRNF
,
1016 .hook
= br_nf_forward_ip
,
1017 .owner
= THIS_MODULE
,
1018 .pf
= NFPROTO_BRIDGE
,
1019 .hooknum
= NF_BR_FORWARD
,
1020 .priority
= NF_BR_PRI_BRNF
- 1,
1023 .hook
= br_nf_forward_arp
,
1024 .owner
= THIS_MODULE
,
1025 .pf
= NFPROTO_BRIDGE
,
1026 .hooknum
= NF_BR_FORWARD
,
1027 .priority
= NF_BR_PRI_BRNF
,
1030 .hook
= br_nf_post_routing
,
1031 .owner
= THIS_MODULE
,
1032 .pf
= NFPROTO_BRIDGE
,
1033 .hooknum
= NF_BR_POST_ROUTING
,
1034 .priority
= NF_BR_PRI_LAST
,
1037 .hook
= ip_sabotage_in
,
1038 .owner
= THIS_MODULE
,
1040 .hooknum
= NF_INET_PRE_ROUTING
,
1041 .priority
= NF_IP_PRI_FIRST
,
1044 .hook
= ip_sabotage_in
,
1045 .owner
= THIS_MODULE
,
1047 .hooknum
= NF_INET_PRE_ROUTING
,
1048 .priority
= NF_IP6_PRI_FIRST
,
1052 #ifdef CONFIG_SYSCTL
1054 int brnf_sysctl_call_tables(struct ctl_table
*ctl
, int write
,
1055 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
1059 ret
= proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
1061 if (write
&& *(int *)(ctl
->data
))
1062 *(int *)(ctl
->data
) = 1;
1066 static struct ctl_table brnf_table
[] = {
1068 .procname
= "bridge-nf-call-arptables",
1069 .data
= &brnf_call_arptables
,
1070 .maxlen
= sizeof(int),
1072 .proc_handler
= brnf_sysctl_call_tables
,
1075 .procname
= "bridge-nf-call-iptables",
1076 .data
= &brnf_call_iptables
,
1077 .maxlen
= sizeof(int),
1079 .proc_handler
= brnf_sysctl_call_tables
,
1082 .procname
= "bridge-nf-call-ip6tables",
1083 .data
= &brnf_call_ip6tables
,
1084 .maxlen
= sizeof(int),
1086 .proc_handler
= brnf_sysctl_call_tables
,
1089 .procname
= "bridge-nf-filter-vlan-tagged",
1090 .data
= &brnf_filter_vlan_tagged
,
1091 .maxlen
= sizeof(int),
1093 .proc_handler
= brnf_sysctl_call_tables
,
1096 .procname
= "bridge-nf-filter-pppoe-tagged",
1097 .data
= &brnf_filter_pppoe_tagged
,
1098 .maxlen
= sizeof(int),
1100 .proc_handler
= brnf_sysctl_call_tables
,
1103 .procname
= "bridge-nf-pass-vlan-input-dev",
1104 .data
= &brnf_pass_vlan_indev
,
1105 .maxlen
= sizeof(int),
1107 .proc_handler
= brnf_sysctl_call_tables
,
1113 static int __init
br_netfilter_init(void)
1117 ret
= nf_register_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
1121 #ifdef CONFIG_SYSCTL
1122 brnf_sysctl_header
= register_net_sysctl(&init_net
, "net/bridge", brnf_table
);
1123 if (brnf_sysctl_header
== NULL
) {
1125 "br_netfilter: can't register to sysctl.\n");
1126 nf_unregister_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
1130 RCU_INIT_POINTER(nf_br_ops
, &br_ops
);
1131 printk(KERN_NOTICE
"Bridge firewalling registered\n");
1135 static void __exit
br_netfilter_fini(void)
1137 RCU_INIT_POINTER(nf_br_ops
, NULL
);
1138 nf_unregister_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
1139 #ifdef CONFIG_SYSCTL
1140 unregister_net_sysctl_table(brnf_sysctl_header
);
1144 module_init(br_netfilter_init
);
1145 module_exit(br_netfilter_fini
);
1147 MODULE_LICENSE("GPL");
1148 MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
1149 MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
1150 MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");