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>
40 #include <net/netns/generic.h>
42 #include <asm/uaccess.h>
43 #include "br_private.h"
45 #include <linux/sysctl.h>
48 static int brnf_net_id __read_mostly
;
55 static struct ctl_table_header
*brnf_sysctl_header
;
56 static int brnf_call_iptables __read_mostly
= 1;
57 static int brnf_call_ip6tables __read_mostly
= 1;
58 static int brnf_call_arptables __read_mostly
= 1;
59 static int brnf_filter_vlan_tagged __read_mostly
;
60 static int brnf_filter_pppoe_tagged __read_mostly
;
61 static int brnf_pass_vlan_indev __read_mostly
;
63 #define brnf_call_iptables 1
64 #define brnf_call_ip6tables 1
65 #define brnf_call_arptables 1
66 #define brnf_filter_vlan_tagged 0
67 #define brnf_filter_pppoe_tagged 0
68 #define brnf_pass_vlan_indev 0
72 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP))
74 #define IS_IPV6(skb) \
75 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6))
78 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP))
80 static inline __be16
vlan_proto(const struct sk_buff
*skb
)
82 if (skb_vlan_tag_present(skb
))
84 else if (skb
->protocol
== htons(ETH_P_8021Q
))
85 return vlan_eth_hdr(skb
)->h_vlan_encapsulated_proto
;
90 #define IS_VLAN_IP(skb) \
91 (vlan_proto(skb) == htons(ETH_P_IP) && \
92 brnf_filter_vlan_tagged)
94 #define IS_VLAN_IPV6(skb) \
95 (vlan_proto(skb) == htons(ETH_P_IPV6) && \
96 brnf_filter_vlan_tagged)
98 #define IS_VLAN_ARP(skb) \
99 (vlan_proto(skb) == htons(ETH_P_ARP) && \
100 brnf_filter_vlan_tagged)
102 static inline __be16
pppoe_proto(const struct sk_buff
*skb
)
104 return *((__be16
*)(skb_mac_header(skb
) + ETH_HLEN
+
105 sizeof(struct pppoe_hdr
)));
108 #define IS_PPPOE_IP(skb) \
109 (skb->protocol == htons(ETH_P_PPP_SES) && \
110 pppoe_proto(skb) == htons(PPP_IP) && \
111 brnf_filter_pppoe_tagged)
113 #define IS_PPPOE_IPV6(skb) \
114 (skb->protocol == htons(ETH_P_PPP_SES) && \
115 pppoe_proto(skb) == htons(PPP_IPV6) && \
116 brnf_filter_pppoe_tagged)
118 /* largest possible L2 header, see br_nf_dev_queue_xmit() */
119 #define NF_BRIDGE_MAX_MAC_HEADER_LENGTH (PPPOE_SES_HLEN + ETH_HLEN)
121 struct brnf_frag_data
{
122 char mac
[NF_BRIDGE_MAX_MAC_HEADER_LENGTH
];
129 static DEFINE_PER_CPU(struct brnf_frag_data
, brnf_frag_data_storage
);
131 static void nf_bridge_info_free(struct sk_buff
*skb
)
133 if (skb
->nf_bridge
) {
134 nf_bridge_put(skb
->nf_bridge
);
135 skb
->nf_bridge
= NULL
;
139 static inline struct net_device
*bridge_parent(const struct net_device
*dev
)
141 struct net_bridge_port
*port
;
143 port
= br_port_get_rcu(dev
);
144 return port
? port
->br
->dev
: NULL
;
147 static inline struct nf_bridge_info
*nf_bridge_unshare(struct sk_buff
*skb
)
149 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
151 if (atomic_read(&nf_bridge
->use
) > 1) {
152 struct nf_bridge_info
*tmp
= nf_bridge_alloc(skb
);
155 memcpy(tmp
, nf_bridge
, sizeof(struct nf_bridge_info
));
156 atomic_set(&tmp
->use
, 1);
158 nf_bridge_put(nf_bridge
);
164 unsigned int nf_bridge_encap_header_len(const struct sk_buff
*skb
)
166 switch (skb
->protocol
) {
167 case __cpu_to_be16(ETH_P_8021Q
):
169 case __cpu_to_be16(ETH_P_PPP_SES
):
170 return PPPOE_SES_HLEN
;
176 static inline void nf_bridge_pull_encap_header(struct sk_buff
*skb
)
178 unsigned int len
= nf_bridge_encap_header_len(skb
);
181 skb
->network_header
+= len
;
184 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff
*skb
)
186 unsigned int len
= nf_bridge_encap_header_len(skb
);
188 skb_pull_rcsum(skb
, len
);
189 skb
->network_header
+= len
;
192 /* When handing a packet over to the IP layer
193 * check whether we have a skb that is in the
197 static int br_validate_ipv4(struct net
*net
, struct sk_buff
*skb
)
199 const struct iphdr
*iph
;
202 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
207 /* Basic sanity checks */
208 if (iph
->ihl
< 5 || iph
->version
!= 4)
211 if (!pskb_may_pull(skb
, iph
->ihl
*4))
215 if (unlikely(ip_fast_csum((u8
*)iph
, iph
->ihl
)))
218 len
= ntohs(iph
->tot_len
);
219 if (skb
->len
< len
) {
220 IP_INC_STATS_BH(net
, IPSTATS_MIB_INTRUNCATEDPKTS
);
222 } else if (len
< (iph
->ihl
*4))
225 if (pskb_trim_rcsum(skb
, len
)) {
226 IP_INC_STATS_BH(net
, IPSTATS_MIB_INDISCARDS
);
230 memset(IPCB(skb
), 0, sizeof(struct inet_skb_parm
));
231 /* We should really parse IP options here but until
232 * somebody who actually uses IP options complains to
233 * us we'll just silently ignore the options because
239 IP_INC_STATS_BH(net
, IPSTATS_MIB_INHDRERRORS
);
244 void nf_bridge_update_protocol(struct sk_buff
*skb
)
246 switch (skb
->nf_bridge
->orig_proto
) {
247 case BRNF_PROTO_8021Q
:
248 skb
->protocol
= htons(ETH_P_8021Q
);
250 case BRNF_PROTO_PPPOE
:
251 skb
->protocol
= htons(ETH_P_PPP_SES
);
253 case BRNF_PROTO_UNCHANGED
:
258 /* Obtain the correct destination MAC address, while preserving the original
259 * source MAC address. If we already know this address, we just copy it. If we
260 * don't, we use the neighbour framework to find out. In both cases, we make
261 * sure that br_handle_frame_finish() is called afterwards.
263 int br_nf_pre_routing_finish_bridge(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
265 struct neighbour
*neigh
;
266 struct dst_entry
*dst
;
268 skb
->dev
= bridge_parent(skb
->dev
);
272 neigh
= dst_neigh_lookup_skb(dst
, skb
);
274 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
277 if (neigh
->hh
.hh_len
) {
278 neigh_hh_bridge(&neigh
->hh
, skb
);
279 skb
->dev
= nf_bridge
->physindev
;
280 ret
= br_handle_frame_finish(net
, sk
, skb
);
282 /* the neighbour function below overwrites the complete
283 * MAC header, so we save the Ethernet source address and
286 skb_copy_from_linear_data_offset(skb
,
287 -(ETH_HLEN
-ETH_ALEN
),
288 nf_bridge
->neigh_header
,
290 /* tell br_dev_xmit to continue with forwarding */
291 nf_bridge
->bridged_dnat
= 1;
292 /* FIXME Need to refragment */
293 ret
= neigh
->output(neigh
, skb
);
295 neigh_release(neigh
);
304 br_nf_ipv4_daddr_was_changed(const struct sk_buff
*skb
,
305 const struct nf_bridge_info
*nf_bridge
)
307 return ip_hdr(skb
)->daddr
!= nf_bridge
->ipv4_daddr
;
310 /* This requires some explaining. If DNAT has taken place,
311 * we will need to fix up the destination Ethernet address.
312 * This is also true when SNAT takes place (for the reply direction).
314 * There are two cases to consider:
315 * 1. The packet was DNAT'ed to a device in the same bridge
316 * port group as it was received on. We can still bridge
318 * 2. The packet was DNAT'ed to a different device, either
319 * a non-bridged device or another bridge port group.
320 * The packet will need to be routed.
322 * The correct way of distinguishing between these two cases is to
323 * call ip_route_input() and to look at skb->dst->dev, which is
324 * changed to the destination device if ip_route_input() succeeds.
326 * Let's first consider the case that ip_route_input() succeeds:
328 * If the output device equals the logical bridge device the packet
329 * came in on, we can consider this bridging. The corresponding MAC
330 * address will be obtained in br_nf_pre_routing_finish_bridge.
331 * Otherwise, the packet is considered to be routed and we just
332 * change the destination MAC address so that the packet will
333 * later be passed up to the IP stack to be routed. For a redirected
334 * packet, ip_route_input() will give back the localhost as output device,
335 * which differs from the bridge device.
337 * Let's now consider the case that ip_route_input() fails:
339 * This can be because the destination address is martian, in which case
340 * the packet will be dropped.
341 * If IP forwarding is disabled, ip_route_input() will fail, while
342 * ip_route_output_key() can return success. The source
343 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
344 * thinks we're handling a locally generated packet and won't care
345 * if IP forwarding is enabled. If the output device equals the logical bridge
346 * device, we proceed as if ip_route_input() succeeded. If it differs from the
347 * logical bridge port or if ip_route_output_key() fails we drop the packet.
349 static int br_nf_pre_routing_finish(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
351 struct net_device
*dev
= skb
->dev
;
352 struct iphdr
*iph
= ip_hdr(skb
);
353 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
357 nf_bridge
->frag_max_size
= IPCB(skb
)->frag_max_size
;
359 if (nf_bridge
->pkt_otherhost
) {
360 skb
->pkt_type
= PACKET_OTHERHOST
;
361 nf_bridge
->pkt_otherhost
= false;
363 nf_bridge
->in_prerouting
= 0;
364 if (br_nf_ipv4_daddr_was_changed(skb
, nf_bridge
)) {
365 if ((err
= ip_route_input(skb
, iph
->daddr
, iph
->saddr
, iph
->tos
, dev
))) {
366 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
368 /* If err equals -EHOSTUNREACH the error is due to a
369 * martian destination or due to the fact that
370 * forwarding is disabled. For most martian packets,
371 * ip_route_output_key() will fail. It won't fail for 2 types of
372 * martian destinations: loopback destinations and destination
373 * 0.0.0.0. In both cases the packet will be dropped because the
374 * destination is the loopback device and not the bridge. */
375 if (err
!= -EHOSTUNREACH
|| !in_dev
|| IN_DEV_FORWARD(in_dev
))
378 rt
= ip_route_output(net
, iph
->daddr
, 0,
379 RT_TOS(iph
->tos
), 0);
381 /* - Bridged-and-DNAT'ed traffic doesn't
382 * require ip_forwarding. */
383 if (rt
->dst
.dev
== dev
) {
384 skb_dst_set(skb
, &rt
->dst
);
393 if (skb_dst(skb
)->dev
== dev
) {
395 skb
->dev
= nf_bridge
->physindev
;
396 nf_bridge_update_protocol(skb
);
397 nf_bridge_push_encap_header(skb
);
398 NF_HOOK_THRESH(NFPROTO_BRIDGE
,
400 net
, sk
, skb
, skb
->dev
, NULL
,
401 br_nf_pre_routing_finish_bridge
,
405 ether_addr_copy(eth_hdr(skb
)->h_dest
, dev
->dev_addr
);
406 skb
->pkt_type
= PACKET_HOST
;
409 rt
= bridge_parent_rtable(nf_bridge
->physindev
);
414 skb_dst_set_noref(skb
, &rt
->dst
);
417 skb
->dev
= nf_bridge
->physindev
;
418 nf_bridge_update_protocol(skb
);
419 nf_bridge_push_encap_header(skb
);
420 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_PRE_ROUTING
, net
, sk
, skb
,
422 br_handle_frame_finish
, 1);
427 static struct net_device
*brnf_get_logical_dev(struct sk_buff
*skb
, const struct net_device
*dev
)
429 struct net_device
*vlan
, *br
;
431 br
= bridge_parent(dev
);
432 if (brnf_pass_vlan_indev
== 0 || !skb_vlan_tag_present(skb
))
435 vlan
= __vlan_find_dev_deep_rcu(br
, skb
->vlan_proto
,
436 skb_vlan_tag_get(skb
) & VLAN_VID_MASK
);
438 return vlan
? vlan
: br
;
441 /* Some common code for IPv4/IPv6 */
442 struct net_device
*setup_pre_routing(struct sk_buff
*skb
)
444 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
446 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
447 skb
->pkt_type
= PACKET_HOST
;
448 nf_bridge
->pkt_otherhost
= true;
451 nf_bridge
->in_prerouting
= 1;
452 nf_bridge
->physindev
= skb
->dev
;
453 skb
->dev
= brnf_get_logical_dev(skb
, skb
->dev
);
455 if (skb
->protocol
== htons(ETH_P_8021Q
))
456 nf_bridge
->orig_proto
= BRNF_PROTO_8021Q
;
457 else if (skb
->protocol
== htons(ETH_P_PPP_SES
))
458 nf_bridge
->orig_proto
= BRNF_PROTO_PPPOE
;
460 /* Must drop socket now because of tproxy. */
465 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
466 * Replicate the checks that IPv4 does on packet reception.
467 * Set skb->dev to the bridge device (i.e. parent of the
468 * receiving device) to make netfilter happy, the REDIRECT
469 * target in particular. Save the original destination IP
470 * address to be able to detect DNAT afterwards. */
471 static unsigned int br_nf_pre_routing(void *priv
,
473 const struct nf_hook_state
*state
)
475 struct nf_bridge_info
*nf_bridge
;
476 struct net_bridge_port
*p
;
477 struct net_bridge
*br
;
478 __u32 len
= nf_bridge_encap_header_len(skb
);
480 if (unlikely(!pskb_may_pull(skb
, len
)))
483 p
= br_port_get_rcu(state
->in
);
488 if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
)) {
489 if (!brnf_call_ip6tables
&& !br
->nf_call_ip6tables
)
492 nf_bridge_pull_encap_header_rcsum(skb
);
493 return br_nf_pre_routing_ipv6(priv
, skb
, state
);
496 if (!brnf_call_iptables
&& !br
->nf_call_iptables
)
499 if (!IS_IP(skb
) && !IS_VLAN_IP(skb
) && !IS_PPPOE_IP(skb
))
502 nf_bridge_pull_encap_header_rcsum(skb
);
504 if (br_validate_ipv4(state
->net
, skb
))
507 nf_bridge_put(skb
->nf_bridge
);
508 if (!nf_bridge_alloc(skb
))
510 if (!setup_pre_routing(skb
))
513 nf_bridge
= nf_bridge_info_get(skb
);
514 nf_bridge
->ipv4_daddr
= ip_hdr(skb
)->daddr
;
516 skb
->protocol
= htons(ETH_P_IP
);
518 NF_HOOK(NFPROTO_IPV4
, NF_INET_PRE_ROUTING
, state
->net
, state
->sk
, skb
,
520 br_nf_pre_routing_finish
);
526 /* PF_BRIDGE/LOCAL_IN ************************************************/
527 /* The packet is locally destined, which requires a real
528 * dst_entry, so detach the fake one. On the way up, the
529 * packet would pass through PRE_ROUTING again (which already
530 * took place when the packet entered the bridge), but we
531 * register an IPv4 PRE_ROUTING 'sabotage' hook that will
532 * prevent this from happening. */
533 static unsigned int br_nf_local_in(void *priv
,
535 const struct nf_hook_state
*state
)
537 br_drop_fake_rtable(skb
);
541 /* PF_BRIDGE/FORWARD *************************************************/
542 static int br_nf_forward_finish(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
544 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
545 struct net_device
*in
;
547 if (!IS_ARP(skb
) && !IS_VLAN_ARP(skb
)) {
549 if (skb
->protocol
== htons(ETH_P_IP
))
550 nf_bridge
->frag_max_size
= IPCB(skb
)->frag_max_size
;
552 if (skb
->protocol
== htons(ETH_P_IPV6
))
553 nf_bridge
->frag_max_size
= IP6CB(skb
)->frag_max_size
;
555 in
= nf_bridge
->physindev
;
556 if (nf_bridge
->pkt_otherhost
) {
557 skb
->pkt_type
= PACKET_OTHERHOST
;
558 nf_bridge
->pkt_otherhost
= false;
560 nf_bridge_update_protocol(skb
);
562 in
= *((struct net_device
**)(skb
->cb
));
564 nf_bridge_push_encap_header(skb
);
566 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_FORWARD
, net
, sk
, skb
,
567 in
, skb
->dev
, br_forward_finish
, 1);
572 /* This is the 'purely bridged' case. For IP, we pass the packet to
573 * netfilter with indev and outdev set to the bridge device,
574 * but we are still able to filter on the 'real' indev/outdev
575 * because of the physdev module. For ARP, indev and outdev are the
577 static unsigned int br_nf_forward_ip(void *priv
,
579 const struct nf_hook_state
*state
)
581 struct nf_bridge_info
*nf_bridge
;
582 struct net_device
*parent
;
588 /* Need exclusive nf_bridge_info since we might have multiple
589 * different physoutdevs. */
590 if (!nf_bridge_unshare(skb
))
593 nf_bridge
= nf_bridge_info_get(skb
);
597 parent
= bridge_parent(state
->out
);
601 if (IS_IP(skb
) || IS_VLAN_IP(skb
) || IS_PPPOE_IP(skb
))
603 else if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
))
608 nf_bridge_pull_encap_header(skb
);
610 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
611 skb
->pkt_type
= PACKET_HOST
;
612 nf_bridge
->pkt_otherhost
= true;
615 if (pf
== NFPROTO_IPV4
) {
616 if (br_validate_ipv4(state
->net
, skb
))
618 IPCB(skb
)->frag_max_size
= nf_bridge
->frag_max_size
;
621 if (pf
== NFPROTO_IPV6
) {
622 if (br_validate_ipv6(state
->net
, skb
))
624 IP6CB(skb
)->frag_max_size
= nf_bridge
->frag_max_size
;
627 nf_bridge
->physoutdev
= skb
->dev
;
628 if (pf
== NFPROTO_IPV4
)
629 skb
->protocol
= htons(ETH_P_IP
);
631 skb
->protocol
= htons(ETH_P_IPV6
);
633 NF_HOOK(pf
, NF_INET_FORWARD
, state
->net
, NULL
, skb
,
634 brnf_get_logical_dev(skb
, state
->in
),
635 parent
, br_nf_forward_finish
);
640 static unsigned int br_nf_forward_arp(void *priv
,
642 const struct nf_hook_state
*state
)
644 struct net_bridge_port
*p
;
645 struct net_bridge
*br
;
646 struct net_device
**d
= (struct net_device
**)(skb
->cb
);
648 p
= br_port_get_rcu(state
->out
);
653 if (!brnf_call_arptables
&& !br
->nf_call_arptables
)
657 if (!IS_VLAN_ARP(skb
))
659 nf_bridge_pull_encap_header(skb
);
662 if (arp_hdr(skb
)->ar_pln
!= 4) {
663 if (IS_VLAN_ARP(skb
))
664 nf_bridge_push_encap_header(skb
);
668 NF_HOOK(NFPROTO_ARP
, NF_ARP_FORWARD
, state
->net
, state
->sk
, skb
,
669 state
->in
, state
->out
, br_nf_forward_finish
);
674 static int br_nf_push_frag_xmit(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
676 struct brnf_frag_data
*data
;
679 data
= this_cpu_ptr(&brnf_frag_data_storage
);
680 err
= skb_cow_head(skb
, data
->size
);
687 if (data
->vlan_tci
) {
688 skb
->vlan_tci
= data
->vlan_tci
;
689 skb
->vlan_proto
= data
->vlan_proto
;
692 skb_copy_to_linear_data_offset(skb
, -data
->size
, data
->mac
, data
->size
);
693 __skb_push(skb
, data
->encap_size
);
695 nf_bridge_info_free(skb
);
696 return br_dev_queue_push_xmit(net
, sk
, skb
);
700 br_nf_ip_fragment(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
,
701 int (*output
)(struct net
*, struct sock
*, struct sk_buff
*))
703 unsigned int mtu
= ip_skb_dst_mtu(skb
);
704 struct iphdr
*iph
= ip_hdr(skb
);
706 if (unlikely(((iph
->frag_off
& htons(IP_DF
)) && !skb
->ignore_df
) ||
707 (IPCB(skb
)->frag_max_size
&&
708 IPCB(skb
)->frag_max_size
> mtu
))) {
709 IP_INC_STATS(net
, IPSTATS_MIB_FRAGFAILS
);
714 return ip_do_fragment(net
, sk
, skb
, output
);
717 static unsigned int nf_bridge_mtu_reduction(const struct sk_buff
*skb
)
719 if (skb
->nf_bridge
->orig_proto
== BRNF_PROTO_PPPOE
)
720 return PPPOE_SES_HLEN
;
724 static int br_nf_dev_queue_xmit(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
726 struct nf_bridge_info
*nf_bridge
;
727 unsigned int mtu_reserved
;
729 mtu_reserved
= nf_bridge_mtu_reduction(skb
);
731 if (skb_is_gso(skb
) || skb
->len
+ mtu_reserved
<= skb
->dev
->mtu
) {
732 nf_bridge_info_free(skb
);
733 return br_dev_queue_push_xmit(net
, sk
, skb
);
736 nf_bridge
= nf_bridge_info_get(skb
);
738 /* This is wrong! We should preserve the original fragment
739 * boundaries by preserving frag_list rather than refragmenting.
741 if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV4
) &&
742 skb
->protocol
== htons(ETH_P_IP
)) {
743 struct brnf_frag_data
*data
;
745 if (br_validate_ipv4(net
, skb
))
748 IPCB(skb
)->frag_max_size
= nf_bridge
->frag_max_size
;
750 nf_bridge_update_protocol(skb
);
752 data
= this_cpu_ptr(&brnf_frag_data_storage
);
754 data
->vlan_tci
= skb
->vlan_tci
;
755 data
->vlan_proto
= skb
->vlan_proto
;
756 data
->encap_size
= nf_bridge_encap_header_len(skb
);
757 data
->size
= ETH_HLEN
+ data
->encap_size
;
759 skb_copy_from_linear_data_offset(skb
, -data
->size
, data
->mac
,
762 return br_nf_ip_fragment(net
, sk
, skb
, br_nf_push_frag_xmit
);
764 if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV6
) &&
765 skb
->protocol
== htons(ETH_P_IPV6
)) {
766 const struct nf_ipv6_ops
*v6ops
= nf_get_ipv6_ops();
767 struct brnf_frag_data
*data
;
769 if (br_validate_ipv6(net
, skb
))
772 IP6CB(skb
)->frag_max_size
= nf_bridge
->frag_max_size
;
774 nf_bridge_update_protocol(skb
);
776 data
= this_cpu_ptr(&brnf_frag_data_storage
);
777 data
->encap_size
= nf_bridge_encap_header_len(skb
);
778 data
->size
= ETH_HLEN
+ data
->encap_size
;
780 skb_copy_from_linear_data_offset(skb
, -data
->size
, data
->mac
,
784 return v6ops
->fragment(net
, sk
, skb
, br_nf_push_frag_xmit
);
789 nf_bridge_info_free(skb
);
790 return br_dev_queue_push_xmit(net
, sk
, skb
);
796 /* PF_BRIDGE/POST_ROUTING ********************************************/
797 static unsigned int br_nf_post_routing(void *priv
,
799 const struct nf_hook_state
*state
)
801 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
802 struct net_device
*realoutdev
= bridge_parent(skb
->dev
);
805 /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
806 * on a bridge, but was delivered locally and is now being routed:
808 * POST_ROUTING was already invoked from the ip stack.
810 if (!nf_bridge
|| !nf_bridge
->physoutdev
)
816 if (IS_IP(skb
) || IS_VLAN_IP(skb
) || IS_PPPOE_IP(skb
))
818 else if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
))
823 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
824 * about the value of skb->pkt_type. */
825 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
826 skb
->pkt_type
= PACKET_HOST
;
827 nf_bridge
->pkt_otherhost
= true;
830 nf_bridge_pull_encap_header(skb
);
831 if (pf
== NFPROTO_IPV4
)
832 skb
->protocol
= htons(ETH_P_IP
);
834 skb
->protocol
= htons(ETH_P_IPV6
);
836 NF_HOOK(pf
, NF_INET_POST_ROUTING
, state
->net
, state
->sk
, skb
,
838 br_nf_dev_queue_xmit
);
843 /* IP/SABOTAGE *****************************************************/
844 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
845 * for the second time. */
846 static unsigned int ip_sabotage_in(void *priv
,
848 const struct nf_hook_state
*state
)
850 if (skb
->nf_bridge
&& !skb
->nf_bridge
->in_prerouting
)
856 /* This is called when br_netfilter has called into iptables/netfilter,
857 * and DNAT has taken place on a bridge-forwarded packet.
859 * neigh->output has created a new MAC header, with local br0 MAC
862 * This restores the original MAC saddr of the bridged packet
863 * before invoking bridge forward logic to transmit the packet.
865 static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff
*skb
)
867 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
869 skb_pull(skb
, ETH_HLEN
);
870 nf_bridge
->bridged_dnat
= 0;
872 BUILD_BUG_ON(sizeof(nf_bridge
->neigh_header
) != (ETH_HLEN
- ETH_ALEN
));
874 skb_copy_to_linear_data_offset(skb
, -(ETH_HLEN
- ETH_ALEN
),
875 nf_bridge
->neigh_header
,
876 ETH_HLEN
- ETH_ALEN
);
877 skb
->dev
= nf_bridge
->physindev
;
879 nf_bridge
->physoutdev
= NULL
;
880 br_handle_frame_finish(dev_net(skb
->dev
), NULL
, skb
);
883 static int br_nf_dev_xmit(struct sk_buff
*skb
)
885 if (skb
->nf_bridge
&& skb
->nf_bridge
->bridged_dnat
) {
886 br_nf_pre_routing_finish_bridge_slow(skb
);
892 static const struct nf_br_ops br_ops
= {
893 .br_dev_xmit_hook
= br_nf_dev_xmit
,
896 void br_netfilter_enable(void)
899 EXPORT_SYMBOL_GPL(br_netfilter_enable
);
901 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
902 * br_dev_queue_push_xmit is called afterwards */
903 static struct nf_hook_ops br_nf_ops
[] __read_mostly
= {
905 .hook
= br_nf_pre_routing
,
906 .pf
= NFPROTO_BRIDGE
,
907 .hooknum
= NF_BR_PRE_ROUTING
,
908 .priority
= NF_BR_PRI_BRNF
,
911 .hook
= br_nf_local_in
,
912 .pf
= NFPROTO_BRIDGE
,
913 .hooknum
= NF_BR_LOCAL_IN
,
914 .priority
= NF_BR_PRI_BRNF
,
917 .hook
= br_nf_forward_ip
,
918 .pf
= NFPROTO_BRIDGE
,
919 .hooknum
= NF_BR_FORWARD
,
920 .priority
= NF_BR_PRI_BRNF
- 1,
923 .hook
= br_nf_forward_arp
,
924 .pf
= NFPROTO_BRIDGE
,
925 .hooknum
= NF_BR_FORWARD
,
926 .priority
= NF_BR_PRI_BRNF
,
929 .hook
= br_nf_post_routing
,
930 .pf
= NFPROTO_BRIDGE
,
931 .hooknum
= NF_BR_POST_ROUTING
,
932 .priority
= NF_BR_PRI_LAST
,
935 .hook
= ip_sabotage_in
,
937 .hooknum
= NF_INET_PRE_ROUTING
,
938 .priority
= NF_IP_PRI_FIRST
,
941 .hook
= ip_sabotage_in
,
943 .hooknum
= NF_INET_PRE_ROUTING
,
944 .priority
= NF_IP6_PRI_FIRST
,
948 static int brnf_device_event(struct notifier_block
*unused
, unsigned long event
,
951 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
952 struct brnf_net
*brnet
;
956 if (event
!= NETDEV_REGISTER
|| !(dev
->priv_flags
& IFF_EBRIDGE
))
962 brnet
= net_generic(net
, brnf_net_id
);
966 ret
= nf_register_net_hooks(net
, br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
970 brnet
->enabled
= true;
974 static void __net_exit
brnf_exit_net(struct net
*net
)
976 struct brnf_net
*brnet
= net_generic(net
, brnf_net_id
);
981 nf_unregister_net_hooks(net
, br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
982 brnet
->enabled
= false;
985 static struct pernet_operations brnf_net_ops __read_mostly
= {
986 .exit
= brnf_exit_net
,
988 .size
= sizeof(struct brnf_net
),
991 static struct notifier_block brnf_notifier __read_mostly
= {
992 .notifier_call
= brnf_device_event
,
997 int brnf_sysctl_call_tables(struct ctl_table
*ctl
, int write
,
998 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
1002 ret
= proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
1004 if (write
&& *(int *)(ctl
->data
))
1005 *(int *)(ctl
->data
) = 1;
1009 static struct ctl_table brnf_table
[] = {
1011 .procname
= "bridge-nf-call-arptables",
1012 .data
= &brnf_call_arptables
,
1013 .maxlen
= sizeof(int),
1015 .proc_handler
= brnf_sysctl_call_tables
,
1018 .procname
= "bridge-nf-call-iptables",
1019 .data
= &brnf_call_iptables
,
1020 .maxlen
= sizeof(int),
1022 .proc_handler
= brnf_sysctl_call_tables
,
1025 .procname
= "bridge-nf-call-ip6tables",
1026 .data
= &brnf_call_ip6tables
,
1027 .maxlen
= sizeof(int),
1029 .proc_handler
= brnf_sysctl_call_tables
,
1032 .procname
= "bridge-nf-filter-vlan-tagged",
1033 .data
= &brnf_filter_vlan_tagged
,
1034 .maxlen
= sizeof(int),
1036 .proc_handler
= brnf_sysctl_call_tables
,
1039 .procname
= "bridge-nf-filter-pppoe-tagged",
1040 .data
= &brnf_filter_pppoe_tagged
,
1041 .maxlen
= sizeof(int),
1043 .proc_handler
= brnf_sysctl_call_tables
,
1046 .procname
= "bridge-nf-pass-vlan-input-dev",
1047 .data
= &brnf_pass_vlan_indev
,
1048 .maxlen
= sizeof(int),
1050 .proc_handler
= brnf_sysctl_call_tables
,
1056 static int __init
br_netfilter_init(void)
1060 ret
= register_pernet_subsys(&brnf_net_ops
);
1064 ret
= register_netdevice_notifier(&brnf_notifier
);
1066 unregister_pernet_subsys(&brnf_net_ops
);
1070 #ifdef CONFIG_SYSCTL
1071 brnf_sysctl_header
= register_net_sysctl(&init_net
, "net/bridge", brnf_table
);
1072 if (brnf_sysctl_header
== NULL
) {
1074 "br_netfilter: can't register to sysctl.\n");
1075 unregister_netdevice_notifier(&brnf_notifier
);
1076 unregister_pernet_subsys(&brnf_net_ops
);
1080 RCU_INIT_POINTER(nf_br_ops
, &br_ops
);
1081 printk(KERN_NOTICE
"Bridge firewalling registered\n");
1085 static void __exit
br_netfilter_fini(void)
1087 RCU_INIT_POINTER(nf_br_ops
, NULL
);
1088 unregister_netdevice_notifier(&brnf_notifier
);
1089 unregister_pernet_subsys(&brnf_net_ops
);
1090 #ifdef CONFIG_SYSCTL
1091 unregister_net_sysctl_table(brnf_sysctl_header
);
1095 module_init(br_netfilter_init
);
1096 module_exit(br_netfilter_fini
);
1098 MODULE_LICENSE("GPL");
1099 MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
1100 MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
1101 MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");