1 // SPDX-License-Identifier: GPL-2.0-or-later
4 * Linux ethernet bridge
7 * Lennert Buytenhek <buytenh@gnu.org>
8 * Bart De Schuymer <bdschuym@pandora.be>
10 * Lennert dedicates this file to Kerstin Wurdinger.
13 #include <linux/module.h>
14 #include <linux/kernel.h>
15 #include <linux/slab.h>
17 #include <linux/netdevice.h>
18 #include <linux/skbuff.h>
19 #include <linux/if_arp.h>
20 #include <linux/if_ether.h>
21 #include <linux/if_vlan.h>
22 #include <linux/if_pppox.h>
23 #include <linux/ppp_defs.h>
24 #include <linux/netfilter_bridge.h>
25 #include <uapi/linux/netfilter_bridge.h>
26 #include <linux/netfilter_ipv4.h>
27 #include <linux/netfilter_ipv6.h>
28 #include <linux/netfilter_arp.h>
29 #include <linux/in_route.h>
30 #include <linux/rculist.h>
31 #include <linux/inetdevice.h>
35 #include <net/addrconf.h>
36 #include <net/route.h>
37 #include <net/netfilter/br_netfilter.h>
38 #include <net/netns/generic.h>
40 #include <linux/uaccess.h>
41 #include "br_private.h"
43 #include <linux/sysctl.h>
46 static unsigned int brnf_net_id __read_mostly
;
53 static struct ctl_table_header
*brnf_sysctl_header
;
54 static int brnf_call_iptables __read_mostly
= 1;
55 static int brnf_call_ip6tables __read_mostly
= 1;
56 static int brnf_call_arptables __read_mostly
= 1;
57 static int brnf_filter_vlan_tagged __read_mostly
;
58 static int brnf_filter_pppoe_tagged __read_mostly
;
59 static int brnf_pass_vlan_indev __read_mostly
;
61 #define brnf_call_iptables 1
62 #define brnf_call_ip6tables 1
63 #define brnf_call_arptables 1
64 #define brnf_filter_vlan_tagged 0
65 #define brnf_filter_pppoe_tagged 0
66 #define brnf_pass_vlan_indev 0
70 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP))
72 #define IS_IPV6(skb) \
73 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6))
76 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP))
78 static inline __be16
vlan_proto(const struct sk_buff
*skb
)
80 if (skb_vlan_tag_present(skb
))
82 else if (skb
->protocol
== htons(ETH_P_8021Q
))
83 return vlan_eth_hdr(skb
)->h_vlan_encapsulated_proto
;
88 #define IS_VLAN_IP(skb) \
89 (vlan_proto(skb) == htons(ETH_P_IP) && \
90 brnf_filter_vlan_tagged)
92 #define IS_VLAN_IPV6(skb) \
93 (vlan_proto(skb) == htons(ETH_P_IPV6) && \
94 brnf_filter_vlan_tagged)
96 #define IS_VLAN_ARP(skb) \
97 (vlan_proto(skb) == htons(ETH_P_ARP) && \
98 brnf_filter_vlan_tagged)
100 static inline __be16
pppoe_proto(const struct sk_buff
*skb
)
102 return *((__be16
*)(skb_mac_header(skb
) + ETH_HLEN
+
103 sizeof(struct pppoe_hdr
)));
106 #define IS_PPPOE_IP(skb) \
107 (skb->protocol == htons(ETH_P_PPP_SES) && \
108 pppoe_proto(skb) == htons(PPP_IP) && \
109 brnf_filter_pppoe_tagged)
111 #define IS_PPPOE_IPV6(skb) \
112 (skb->protocol == htons(ETH_P_PPP_SES) && \
113 pppoe_proto(skb) == htons(PPP_IPV6) && \
114 brnf_filter_pppoe_tagged)
116 /* largest possible L2 header, see br_nf_dev_queue_xmit() */
117 #define NF_BRIDGE_MAX_MAC_HEADER_LENGTH (PPPOE_SES_HLEN + ETH_HLEN)
119 struct brnf_frag_data
{
120 char mac
[NF_BRIDGE_MAX_MAC_HEADER_LENGTH
];
127 static DEFINE_PER_CPU(struct brnf_frag_data
, brnf_frag_data_storage
);
129 static void nf_bridge_info_free(struct sk_buff
*skb
)
131 skb_ext_del(skb
, SKB_EXT_BRIDGE_NF
);
134 static inline struct net_device
*bridge_parent(const struct net_device
*dev
)
136 struct net_bridge_port
*port
;
138 port
= br_port_get_rcu(dev
);
139 return port
? port
->br
->dev
: NULL
;
142 static inline struct nf_bridge_info
*nf_bridge_unshare(struct sk_buff
*skb
)
144 return skb_ext_add(skb
, SKB_EXT_BRIDGE_NF
);
147 unsigned int nf_bridge_encap_header_len(const struct sk_buff
*skb
)
149 switch (skb
->protocol
) {
150 case __cpu_to_be16(ETH_P_8021Q
):
152 case __cpu_to_be16(ETH_P_PPP_SES
):
153 return PPPOE_SES_HLEN
;
159 static inline void nf_bridge_pull_encap_header(struct sk_buff
*skb
)
161 unsigned int len
= nf_bridge_encap_header_len(skb
);
164 skb
->network_header
+= len
;
167 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff
*skb
)
169 unsigned int len
= nf_bridge_encap_header_len(skb
);
171 skb_pull_rcsum(skb
, len
);
172 skb
->network_header
+= len
;
175 /* When handing a packet over to the IP layer
176 * check whether we have a skb that is in the
180 static int br_validate_ipv4(struct net
*net
, struct sk_buff
*skb
)
182 const struct iphdr
*iph
;
185 if (!pskb_may_pull(skb
, sizeof(struct iphdr
)))
190 /* Basic sanity checks */
191 if (iph
->ihl
< 5 || iph
->version
!= 4)
194 if (!pskb_may_pull(skb
, iph
->ihl
*4))
198 if (unlikely(ip_fast_csum((u8
*)iph
, iph
->ihl
)))
201 len
= ntohs(iph
->tot_len
);
202 if (skb
->len
< len
) {
203 __IP_INC_STATS(net
, IPSTATS_MIB_INTRUNCATEDPKTS
);
205 } else if (len
< (iph
->ihl
*4))
208 if (pskb_trim_rcsum(skb
, len
)) {
209 __IP_INC_STATS(net
, IPSTATS_MIB_INDISCARDS
);
213 memset(IPCB(skb
), 0, sizeof(struct inet_skb_parm
));
214 /* We should really parse IP options here but until
215 * somebody who actually uses IP options complains to
216 * us we'll just silently ignore the options because
222 __IP_INC_STATS(net
, IPSTATS_MIB_CSUMERRORS
);
224 __IP_INC_STATS(net
, IPSTATS_MIB_INHDRERRORS
);
229 void nf_bridge_update_protocol(struct sk_buff
*skb
)
231 const struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
233 switch (nf_bridge
->orig_proto
) {
234 case BRNF_PROTO_8021Q
:
235 skb
->protocol
= htons(ETH_P_8021Q
);
237 case BRNF_PROTO_PPPOE
:
238 skb
->protocol
= htons(ETH_P_PPP_SES
);
240 case BRNF_PROTO_UNCHANGED
:
245 /* Obtain the correct destination MAC address, while preserving the original
246 * source MAC address. If we already know this address, we just copy it. If we
247 * don't, we use the neighbour framework to find out. In both cases, we make
248 * sure that br_handle_frame_finish() is called afterwards.
250 int br_nf_pre_routing_finish_bridge(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
252 struct neighbour
*neigh
;
253 struct dst_entry
*dst
;
255 skb
->dev
= bridge_parent(skb
->dev
);
259 neigh
= dst_neigh_lookup_skb(dst
, skb
);
261 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
264 if ((neigh
->nud_state
& NUD_CONNECTED
) && neigh
->hh
.hh_len
) {
265 neigh_hh_bridge(&neigh
->hh
, skb
);
266 skb
->dev
= nf_bridge
->physindev
;
267 ret
= br_handle_frame_finish(net
, sk
, skb
);
269 /* the neighbour function below overwrites the complete
270 * MAC header, so we save the Ethernet source address and
273 skb_copy_from_linear_data_offset(skb
,
274 -(ETH_HLEN
-ETH_ALEN
),
275 nf_bridge
->neigh_header
,
277 /* tell br_dev_xmit to continue with forwarding */
278 nf_bridge
->bridged_dnat
= 1;
279 /* FIXME Need to refragment */
280 ret
= neigh
->output(neigh
, skb
);
282 neigh_release(neigh
);
291 br_nf_ipv4_daddr_was_changed(const struct sk_buff
*skb
,
292 const struct nf_bridge_info
*nf_bridge
)
294 return ip_hdr(skb
)->daddr
!= nf_bridge
->ipv4_daddr
;
297 /* This requires some explaining. If DNAT has taken place,
298 * we will need to fix up the destination Ethernet address.
299 * This is also true when SNAT takes place (for the reply direction).
301 * There are two cases to consider:
302 * 1. The packet was DNAT'ed to a device in the same bridge
303 * port group as it was received on. We can still bridge
305 * 2. The packet was DNAT'ed to a different device, either
306 * a non-bridged device or another bridge port group.
307 * The packet will need to be routed.
309 * The correct way of distinguishing between these two cases is to
310 * call ip_route_input() and to look at skb->dst->dev, which is
311 * changed to the destination device if ip_route_input() succeeds.
313 * Let's first consider the case that ip_route_input() succeeds:
315 * If the output device equals the logical bridge device the packet
316 * came in on, we can consider this bridging. The corresponding MAC
317 * address will be obtained in br_nf_pre_routing_finish_bridge.
318 * Otherwise, the packet is considered to be routed and we just
319 * change the destination MAC address so that the packet will
320 * later be passed up to the IP stack to be routed. For a redirected
321 * packet, ip_route_input() will give back the localhost as output device,
322 * which differs from the bridge device.
324 * Let's now consider the case that ip_route_input() fails:
326 * This can be because the destination address is martian, in which case
327 * the packet will be dropped.
328 * If IP forwarding is disabled, ip_route_input() will fail, while
329 * ip_route_output_key() can return success. The source
330 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
331 * thinks we're handling a locally generated packet and won't care
332 * if IP forwarding is enabled. If the output device equals the logical bridge
333 * device, we proceed as if ip_route_input() succeeded. If it differs from the
334 * logical bridge port or if ip_route_output_key() fails we drop the packet.
336 static int br_nf_pre_routing_finish(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
338 struct net_device
*dev
= skb
->dev
;
339 struct iphdr
*iph
= ip_hdr(skb
);
340 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
344 nf_bridge
->frag_max_size
= IPCB(skb
)->frag_max_size
;
346 if (nf_bridge
->pkt_otherhost
) {
347 skb
->pkt_type
= PACKET_OTHERHOST
;
348 nf_bridge
->pkt_otherhost
= false;
350 nf_bridge
->in_prerouting
= 0;
351 if (br_nf_ipv4_daddr_was_changed(skb
, nf_bridge
)) {
352 if ((err
= ip_route_input(skb
, iph
->daddr
, iph
->saddr
, iph
->tos
, dev
))) {
353 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
355 /* If err equals -EHOSTUNREACH the error is due to a
356 * martian destination or due to the fact that
357 * forwarding is disabled. For most martian packets,
358 * ip_route_output_key() will fail. It won't fail for 2 types of
359 * martian destinations: loopback destinations and destination
360 * 0.0.0.0. In both cases the packet will be dropped because the
361 * destination is the loopback device and not the bridge. */
362 if (err
!= -EHOSTUNREACH
|| !in_dev
|| IN_DEV_FORWARD(in_dev
))
365 rt
= ip_route_output(net
, iph
->daddr
, 0,
366 RT_TOS(iph
->tos
), 0);
368 /* - Bridged-and-DNAT'ed traffic doesn't
369 * require ip_forwarding. */
370 if (rt
->dst
.dev
== dev
) {
371 skb_dst_set(skb
, &rt
->dst
);
380 if (skb_dst(skb
)->dev
== dev
) {
382 skb
->dev
= nf_bridge
->physindev
;
383 nf_bridge_update_protocol(skb
);
384 nf_bridge_push_encap_header(skb
);
385 br_nf_hook_thresh(NF_BR_PRE_ROUTING
,
386 net
, sk
, skb
, skb
->dev
,
388 br_nf_pre_routing_finish_bridge
);
391 ether_addr_copy(eth_hdr(skb
)->h_dest
, dev
->dev_addr
);
392 skb
->pkt_type
= PACKET_HOST
;
395 rt
= bridge_parent_rtable(nf_bridge
->physindev
);
400 skb_dst_set_noref(skb
, &rt
->dst
);
403 skb
->dev
= nf_bridge
->physindev
;
404 nf_bridge_update_protocol(skb
);
405 nf_bridge_push_encap_header(skb
);
406 br_nf_hook_thresh(NF_BR_PRE_ROUTING
, net
, sk
, skb
, skb
->dev
, NULL
,
407 br_handle_frame_finish
);
411 static struct net_device
*brnf_get_logical_dev(struct sk_buff
*skb
, const struct net_device
*dev
)
413 struct net_device
*vlan
, *br
;
415 br
= bridge_parent(dev
);
416 if (brnf_pass_vlan_indev
== 0 || !skb_vlan_tag_present(skb
))
419 vlan
= __vlan_find_dev_deep_rcu(br
, skb
->vlan_proto
,
420 skb_vlan_tag_get(skb
) & VLAN_VID_MASK
);
422 return vlan
? vlan
: br
;
425 /* Some common code for IPv4/IPv6 */
426 struct net_device
*setup_pre_routing(struct sk_buff
*skb
)
428 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
430 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
431 skb
->pkt_type
= PACKET_HOST
;
432 nf_bridge
->pkt_otherhost
= true;
435 nf_bridge
->in_prerouting
= 1;
436 nf_bridge
->physindev
= skb
->dev
;
437 skb
->dev
= brnf_get_logical_dev(skb
, skb
->dev
);
439 if (skb
->protocol
== htons(ETH_P_8021Q
))
440 nf_bridge
->orig_proto
= BRNF_PROTO_8021Q
;
441 else if (skb
->protocol
== htons(ETH_P_PPP_SES
))
442 nf_bridge
->orig_proto
= BRNF_PROTO_PPPOE
;
444 /* Must drop socket now because of tproxy. */
449 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
450 * Replicate the checks that IPv4 does on packet reception.
451 * Set skb->dev to the bridge device (i.e. parent of the
452 * receiving device) to make netfilter happy, the REDIRECT
453 * target in particular. Save the original destination IP
454 * address to be able to detect DNAT afterwards. */
455 static unsigned int br_nf_pre_routing(void *priv
,
457 const struct nf_hook_state
*state
)
459 struct nf_bridge_info
*nf_bridge
;
460 struct net_bridge_port
*p
;
461 struct net_bridge
*br
;
462 __u32 len
= nf_bridge_encap_header_len(skb
);
464 if (unlikely(!pskb_may_pull(skb
, len
)))
467 p
= br_port_get_rcu(state
->in
);
472 if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
)) {
473 if (!brnf_call_ip6tables
&&
474 !br_opt_get(br
, BROPT_NF_CALL_IP6TABLES
))
477 nf_bridge_pull_encap_header_rcsum(skb
);
478 return br_nf_pre_routing_ipv6(priv
, skb
, state
);
481 if (!brnf_call_iptables
&& !br_opt_get(br
, BROPT_NF_CALL_IPTABLES
))
484 if (!IS_IP(skb
) && !IS_VLAN_IP(skb
) && !IS_PPPOE_IP(skb
))
487 nf_bridge_pull_encap_header_rcsum(skb
);
489 if (br_validate_ipv4(state
->net
, skb
))
492 if (!nf_bridge_alloc(skb
))
494 if (!setup_pre_routing(skb
))
497 nf_bridge
= nf_bridge_info_get(skb
);
498 nf_bridge
->ipv4_daddr
= ip_hdr(skb
)->daddr
;
500 skb
->protocol
= htons(ETH_P_IP
);
501 skb
->transport_header
= skb
->network_header
+ ip_hdr(skb
)->ihl
* 4;
503 NF_HOOK(NFPROTO_IPV4
, NF_INET_PRE_ROUTING
, state
->net
, state
->sk
, skb
,
505 br_nf_pre_routing_finish
);
511 /* PF_BRIDGE/FORWARD *************************************************/
512 static int br_nf_forward_finish(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
514 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
515 struct net_device
*in
;
517 if (!IS_ARP(skb
) && !IS_VLAN_ARP(skb
)) {
519 if (skb
->protocol
== htons(ETH_P_IP
))
520 nf_bridge
->frag_max_size
= IPCB(skb
)->frag_max_size
;
522 if (skb
->protocol
== htons(ETH_P_IPV6
))
523 nf_bridge
->frag_max_size
= IP6CB(skb
)->frag_max_size
;
525 in
= nf_bridge
->physindev
;
526 if (nf_bridge
->pkt_otherhost
) {
527 skb
->pkt_type
= PACKET_OTHERHOST
;
528 nf_bridge
->pkt_otherhost
= false;
530 nf_bridge_update_protocol(skb
);
532 in
= *((struct net_device
**)(skb
->cb
));
534 nf_bridge_push_encap_header(skb
);
536 br_nf_hook_thresh(NF_BR_FORWARD
, net
, sk
, skb
, in
, skb
->dev
,
542 /* This is the 'purely bridged' case. For IP, we pass the packet to
543 * netfilter with indev and outdev set to the bridge device,
544 * but we are still able to filter on the 'real' indev/outdev
545 * because of the physdev module. For ARP, indev and outdev are the
547 static unsigned int br_nf_forward_ip(void *priv
,
549 const struct nf_hook_state
*state
)
551 struct nf_bridge_info
*nf_bridge
;
552 struct net_device
*parent
;
555 nf_bridge
= nf_bridge_info_get(skb
);
559 /* Need exclusive nf_bridge_info since we might have multiple
560 * different physoutdevs. */
561 if (!nf_bridge_unshare(skb
))
564 nf_bridge
= nf_bridge_info_get(skb
);
568 parent
= bridge_parent(state
->out
);
572 if (IS_IP(skb
) || IS_VLAN_IP(skb
) || IS_PPPOE_IP(skb
))
574 else if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
))
579 nf_bridge_pull_encap_header(skb
);
581 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
582 skb
->pkt_type
= PACKET_HOST
;
583 nf_bridge
->pkt_otherhost
= true;
586 if (pf
== NFPROTO_IPV4
) {
587 if (br_validate_ipv4(state
->net
, skb
))
589 IPCB(skb
)->frag_max_size
= nf_bridge
->frag_max_size
;
592 if (pf
== NFPROTO_IPV6
) {
593 if (br_validate_ipv6(state
->net
, skb
))
595 IP6CB(skb
)->frag_max_size
= nf_bridge
->frag_max_size
;
598 nf_bridge
->physoutdev
= skb
->dev
;
599 if (pf
== NFPROTO_IPV4
)
600 skb
->protocol
= htons(ETH_P_IP
);
602 skb
->protocol
= htons(ETH_P_IPV6
);
604 NF_HOOK(pf
, NF_INET_FORWARD
, state
->net
, NULL
, skb
,
605 brnf_get_logical_dev(skb
, state
->in
),
606 parent
, br_nf_forward_finish
);
611 static unsigned int br_nf_forward_arp(void *priv
,
613 const struct nf_hook_state
*state
)
615 struct net_bridge_port
*p
;
616 struct net_bridge
*br
;
617 struct net_device
**d
= (struct net_device
**)(skb
->cb
);
619 p
= br_port_get_rcu(state
->out
);
624 if (!brnf_call_arptables
&& !br_opt_get(br
, BROPT_NF_CALL_ARPTABLES
))
628 if (!IS_VLAN_ARP(skb
))
630 nf_bridge_pull_encap_header(skb
);
633 if (arp_hdr(skb
)->ar_pln
!= 4) {
634 if (IS_VLAN_ARP(skb
))
635 nf_bridge_push_encap_header(skb
);
639 NF_HOOK(NFPROTO_ARP
, NF_ARP_FORWARD
, state
->net
, state
->sk
, skb
,
640 state
->in
, state
->out
, br_nf_forward_finish
);
645 static int br_nf_push_frag_xmit(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
647 struct brnf_frag_data
*data
;
650 data
= this_cpu_ptr(&brnf_frag_data_storage
);
651 err
= skb_cow_head(skb
, data
->size
);
658 if (data
->vlan_proto
)
659 __vlan_hwaccel_put_tag(skb
, data
->vlan_proto
, data
->vlan_tci
);
661 skb_copy_to_linear_data_offset(skb
, -data
->size
, data
->mac
, data
->size
);
662 __skb_push(skb
, data
->encap_size
);
664 nf_bridge_info_free(skb
);
665 return br_dev_queue_push_xmit(net
, sk
, skb
);
669 br_nf_ip_fragment(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
,
670 int (*output
)(struct net
*, struct sock
*, struct sk_buff
*))
672 unsigned int mtu
= ip_skb_dst_mtu(sk
, skb
);
673 struct iphdr
*iph
= ip_hdr(skb
);
675 if (unlikely(((iph
->frag_off
& htons(IP_DF
)) && !skb
->ignore_df
) ||
676 (IPCB(skb
)->frag_max_size
&&
677 IPCB(skb
)->frag_max_size
> mtu
))) {
678 IP_INC_STATS(net
, IPSTATS_MIB_FRAGFAILS
);
683 return ip_do_fragment(net
, sk
, skb
, output
);
686 static unsigned int nf_bridge_mtu_reduction(const struct sk_buff
*skb
)
688 const struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
690 if (nf_bridge
->orig_proto
== BRNF_PROTO_PPPOE
)
691 return PPPOE_SES_HLEN
;
695 static int br_nf_dev_queue_xmit(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
697 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
698 unsigned int mtu
, mtu_reserved
;
700 mtu_reserved
= nf_bridge_mtu_reduction(skb
);
703 if (nf_bridge
->frag_max_size
&& nf_bridge
->frag_max_size
< mtu
)
704 mtu
= nf_bridge
->frag_max_size
;
706 if (skb_is_gso(skb
) || skb
->len
+ mtu_reserved
<= mtu
) {
707 nf_bridge_info_free(skb
);
708 return br_dev_queue_push_xmit(net
, sk
, skb
);
711 /* This is wrong! We should preserve the original fragment
712 * boundaries by preserving frag_list rather than refragmenting.
714 if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV4
) &&
715 skb
->protocol
== htons(ETH_P_IP
)) {
716 struct brnf_frag_data
*data
;
718 if (br_validate_ipv4(net
, skb
))
721 IPCB(skb
)->frag_max_size
= nf_bridge
->frag_max_size
;
723 nf_bridge_update_protocol(skb
);
725 data
= this_cpu_ptr(&brnf_frag_data_storage
);
727 if (skb_vlan_tag_present(skb
)) {
728 data
->vlan_tci
= skb
->vlan_tci
;
729 data
->vlan_proto
= skb
->vlan_proto
;
731 data
->vlan_proto
= 0;
734 data
->encap_size
= nf_bridge_encap_header_len(skb
);
735 data
->size
= ETH_HLEN
+ data
->encap_size
;
737 skb_copy_from_linear_data_offset(skb
, -data
->size
, data
->mac
,
740 return br_nf_ip_fragment(net
, sk
, skb
, br_nf_push_frag_xmit
);
742 if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV6
) &&
743 skb
->protocol
== htons(ETH_P_IPV6
)) {
744 const struct nf_ipv6_ops
*v6ops
= nf_get_ipv6_ops();
745 struct brnf_frag_data
*data
;
747 if (br_validate_ipv6(net
, skb
))
750 IP6CB(skb
)->frag_max_size
= nf_bridge
->frag_max_size
;
752 nf_bridge_update_protocol(skb
);
754 data
= this_cpu_ptr(&brnf_frag_data_storage
);
755 data
->encap_size
= nf_bridge_encap_header_len(skb
);
756 data
->size
= ETH_HLEN
+ data
->encap_size
;
758 skb_copy_from_linear_data_offset(skb
, -data
->size
, data
->mac
,
762 return v6ops
->fragment(net
, sk
, skb
, br_nf_push_frag_xmit
);
767 nf_bridge_info_free(skb
);
768 return br_dev_queue_push_xmit(net
, sk
, skb
);
774 /* PF_BRIDGE/POST_ROUTING ********************************************/
775 static unsigned int br_nf_post_routing(void *priv
,
777 const struct nf_hook_state
*state
)
779 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
780 struct net_device
*realoutdev
= bridge_parent(skb
->dev
);
783 /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
784 * on a bridge, but was delivered locally and is now being routed:
786 * POST_ROUTING was already invoked from the ip stack.
788 if (!nf_bridge
|| !nf_bridge
->physoutdev
)
794 if (IS_IP(skb
) || IS_VLAN_IP(skb
) || IS_PPPOE_IP(skb
))
796 else if (IS_IPV6(skb
) || IS_VLAN_IPV6(skb
) || IS_PPPOE_IPV6(skb
))
801 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
802 * about the value of skb->pkt_type. */
803 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
804 skb
->pkt_type
= PACKET_HOST
;
805 nf_bridge
->pkt_otherhost
= true;
808 nf_bridge_pull_encap_header(skb
);
809 if (pf
== NFPROTO_IPV4
)
810 skb
->protocol
= htons(ETH_P_IP
);
812 skb
->protocol
= htons(ETH_P_IPV6
);
814 NF_HOOK(pf
, NF_INET_POST_ROUTING
, state
->net
, state
->sk
, skb
,
816 br_nf_dev_queue_xmit
);
821 /* IP/SABOTAGE *****************************************************/
822 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
823 * for the second time. */
824 static unsigned int ip_sabotage_in(void *priv
,
826 const struct nf_hook_state
*state
)
828 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
830 if (nf_bridge
&& !nf_bridge
->in_prerouting
&&
831 !netif_is_l3_master(skb
->dev
) &&
832 !netif_is_l3_slave(skb
->dev
)) {
833 state
->okfn(state
->net
, state
->sk
, skb
);
840 /* This is called when br_netfilter has called into iptables/netfilter,
841 * and DNAT has taken place on a bridge-forwarded packet.
843 * neigh->output has created a new MAC header, with local br0 MAC
846 * This restores the original MAC saddr of the bridged packet
847 * before invoking bridge forward logic to transmit the packet.
849 static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff
*skb
)
851 struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
853 skb_pull(skb
, ETH_HLEN
);
854 nf_bridge
->bridged_dnat
= 0;
856 BUILD_BUG_ON(sizeof(nf_bridge
->neigh_header
) != (ETH_HLEN
- ETH_ALEN
));
858 skb_copy_to_linear_data_offset(skb
, -(ETH_HLEN
- ETH_ALEN
),
859 nf_bridge
->neigh_header
,
860 ETH_HLEN
- ETH_ALEN
);
861 skb
->dev
= nf_bridge
->physindev
;
863 nf_bridge
->physoutdev
= NULL
;
864 br_handle_frame_finish(dev_net(skb
->dev
), NULL
, skb
);
867 static int br_nf_dev_xmit(struct sk_buff
*skb
)
869 const struct nf_bridge_info
*nf_bridge
= nf_bridge_info_get(skb
);
871 if (nf_bridge
&& nf_bridge
->bridged_dnat
) {
872 br_nf_pre_routing_finish_bridge_slow(skb
);
878 static const struct nf_br_ops br_ops
= {
879 .br_dev_xmit_hook
= br_nf_dev_xmit
,
882 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
883 * br_dev_queue_push_xmit is called afterwards */
884 static const struct nf_hook_ops br_nf_ops
[] = {
886 .hook
= br_nf_pre_routing
,
887 .pf
= NFPROTO_BRIDGE
,
888 .hooknum
= NF_BR_PRE_ROUTING
,
889 .priority
= NF_BR_PRI_BRNF
,
892 .hook
= br_nf_forward_ip
,
893 .pf
= NFPROTO_BRIDGE
,
894 .hooknum
= NF_BR_FORWARD
,
895 .priority
= NF_BR_PRI_BRNF
- 1,
898 .hook
= br_nf_forward_arp
,
899 .pf
= NFPROTO_BRIDGE
,
900 .hooknum
= NF_BR_FORWARD
,
901 .priority
= NF_BR_PRI_BRNF
,
904 .hook
= br_nf_post_routing
,
905 .pf
= NFPROTO_BRIDGE
,
906 .hooknum
= NF_BR_POST_ROUTING
,
907 .priority
= NF_BR_PRI_LAST
,
910 .hook
= ip_sabotage_in
,
912 .hooknum
= NF_INET_PRE_ROUTING
,
913 .priority
= NF_IP_PRI_FIRST
,
916 .hook
= ip_sabotage_in
,
918 .hooknum
= NF_INET_PRE_ROUTING
,
919 .priority
= NF_IP6_PRI_FIRST
,
923 static int brnf_device_event(struct notifier_block
*unused
, unsigned long event
,
926 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
927 struct brnf_net
*brnet
;
931 if (event
!= NETDEV_REGISTER
|| !(dev
->priv_flags
& IFF_EBRIDGE
))
937 brnet
= net_generic(net
, brnf_net_id
);
941 ret
= nf_register_net_hooks(net
, br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
945 brnet
->enabled
= true;
949 static void __net_exit
brnf_exit_net(struct net
*net
)
951 struct brnf_net
*brnet
= net_generic(net
, brnf_net_id
);
956 nf_unregister_net_hooks(net
, br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
957 brnet
->enabled
= false;
960 static struct pernet_operations brnf_net_ops __read_mostly
= {
961 .exit
= brnf_exit_net
,
963 .size
= sizeof(struct brnf_net
),
966 static struct notifier_block brnf_notifier __read_mostly
= {
967 .notifier_call
= brnf_device_event
,
970 /* recursively invokes nf_hook_slow (again), skipping already-called
971 * hooks (< NF_BR_PRI_BRNF).
973 * Called with rcu read lock held.
975 int br_nf_hook_thresh(unsigned int hook
, struct net
*net
,
976 struct sock
*sk
, struct sk_buff
*skb
,
977 struct net_device
*indev
,
978 struct net_device
*outdev
,
979 int (*okfn
)(struct net
*, struct sock
*,
982 const struct nf_hook_entries
*e
;
983 struct nf_hook_state state
;
984 struct nf_hook_ops
**ops
;
988 e
= rcu_dereference(net
->nf
.hooks_bridge
[hook
]);
990 return okfn(net
, sk
, skb
);
992 ops
= nf_hook_entries_get_hook_ops(e
);
993 for (i
= 0; i
< e
->num_hook_entries
&&
994 ops
[i
]->priority
<= NF_BR_PRI_BRNF
; i
++)
997 nf_hook_state_init(&state
, hook
, NFPROTO_BRIDGE
, indev
, outdev
,
1000 ret
= nf_hook_slow(skb
, &state
, e
, i
);
1002 ret
= okfn(net
, sk
, skb
);
1007 #ifdef CONFIG_SYSCTL
1009 int brnf_sysctl_call_tables(struct ctl_table
*ctl
, int write
,
1010 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
1014 ret
= proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
1016 if (write
&& *(int *)(ctl
->data
))
1017 *(int *)(ctl
->data
) = 1;
1021 static struct ctl_table brnf_table
[] = {
1023 .procname
= "bridge-nf-call-arptables",
1024 .data
= &brnf_call_arptables
,
1025 .maxlen
= sizeof(int),
1027 .proc_handler
= brnf_sysctl_call_tables
,
1030 .procname
= "bridge-nf-call-iptables",
1031 .data
= &brnf_call_iptables
,
1032 .maxlen
= sizeof(int),
1034 .proc_handler
= brnf_sysctl_call_tables
,
1037 .procname
= "bridge-nf-call-ip6tables",
1038 .data
= &brnf_call_ip6tables
,
1039 .maxlen
= sizeof(int),
1041 .proc_handler
= brnf_sysctl_call_tables
,
1044 .procname
= "bridge-nf-filter-vlan-tagged",
1045 .data
= &brnf_filter_vlan_tagged
,
1046 .maxlen
= sizeof(int),
1048 .proc_handler
= brnf_sysctl_call_tables
,
1051 .procname
= "bridge-nf-filter-pppoe-tagged",
1052 .data
= &brnf_filter_pppoe_tagged
,
1053 .maxlen
= sizeof(int),
1055 .proc_handler
= brnf_sysctl_call_tables
,
1058 .procname
= "bridge-nf-pass-vlan-input-dev",
1059 .data
= &brnf_pass_vlan_indev
,
1060 .maxlen
= sizeof(int),
1062 .proc_handler
= brnf_sysctl_call_tables
,
1068 static int __init
br_netfilter_init(void)
1072 ret
= register_pernet_subsys(&brnf_net_ops
);
1076 ret
= register_netdevice_notifier(&brnf_notifier
);
1078 unregister_pernet_subsys(&brnf_net_ops
);
1082 #ifdef CONFIG_SYSCTL
1083 brnf_sysctl_header
= register_net_sysctl(&init_net
, "net/bridge", brnf_table
);
1084 if (brnf_sysctl_header
== NULL
) {
1086 "br_netfilter: can't register to sysctl.\n");
1087 unregister_netdevice_notifier(&brnf_notifier
);
1088 unregister_pernet_subsys(&brnf_net_ops
);
1092 RCU_INIT_POINTER(nf_br_ops
, &br_ops
);
1093 printk(KERN_NOTICE
"Bridge firewalling registered\n");
1097 static void __exit
br_netfilter_fini(void)
1099 RCU_INIT_POINTER(nf_br_ops
, NULL
);
1100 unregister_netdevice_notifier(&brnf_notifier
);
1101 unregister_pernet_subsys(&brnf_net_ops
);
1102 #ifdef CONFIG_SYSCTL
1103 unregister_net_sysctl_table(brnf_sysctl_header
);
1107 module_init(br_netfilter_init
);
1108 module_exit(br_netfilter_fini
);
1110 MODULE_LICENSE("GPL");
1111 MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
1112 MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
1113 MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");