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>
39 #include <asm/uaccess.h>
40 #include "br_private.h"
42 #include <linux/sysctl.h>
45 #define skb_origaddr(skb) (((struct bridge_skb_cb *) \
46 (skb->nf_bridge->data))->daddr.ipv4)
47 #define store_orig_dstaddr(skb) (skb_origaddr(skb) = ip_hdr(skb)->daddr)
48 #define dnat_took_place(skb) (skb_origaddr(skb) != ip_hdr(skb)->daddr)
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;
58 #define brnf_call_iptables 1
59 #define brnf_call_ip6tables 1
60 #define brnf_call_arptables 1
61 #define brnf_filter_vlan_tagged 0
62 #define brnf_filter_pppoe_tagged 0
65 static inline __be16
vlan_proto(const struct sk_buff
*skb
)
67 if (vlan_tx_tag_present(skb
))
69 else if (skb
->protocol
== htons(ETH_P_8021Q
))
70 return vlan_eth_hdr(skb
)->h_vlan_encapsulated_proto
;
75 #define IS_VLAN_IP(skb) \
76 (vlan_proto(skb) == htons(ETH_P_IP) && \
77 brnf_filter_vlan_tagged)
79 #define IS_VLAN_IPV6(skb) \
80 (vlan_proto(skb) == htons(ETH_P_IPV6) && \
81 brnf_filter_vlan_tagged)
83 #define IS_VLAN_ARP(skb) \
84 (vlan_proto(skb) == htons(ETH_P_ARP) && \
85 brnf_filter_vlan_tagged)
87 static inline __be16
pppoe_proto(const struct sk_buff
*skb
)
89 return *((__be16
*)(skb_mac_header(skb
) + ETH_HLEN
+
90 sizeof(struct pppoe_hdr
)));
93 #define IS_PPPOE_IP(skb) \
94 (skb->protocol == htons(ETH_P_PPP_SES) && \
95 pppoe_proto(skb) == htons(PPP_IP) && \
96 brnf_filter_pppoe_tagged)
98 #define IS_PPPOE_IPV6(skb) \
99 (skb->protocol == htons(ETH_P_PPP_SES) && \
100 pppoe_proto(skb) == htons(PPP_IPV6) && \
101 brnf_filter_pppoe_tagged)
103 static void fake_update_pmtu(struct dst_entry
*dst
, u32 mtu
)
107 static u32
*fake_cow_metrics(struct dst_entry
*dst
, unsigned long old
)
112 static struct dst_ops fake_dst_ops
= {
114 .protocol
= cpu_to_be16(ETH_P_IP
),
115 .update_pmtu
= fake_update_pmtu
,
116 .cow_metrics
= fake_cow_metrics
,
120 * Initialize bogus route table used to keep netfilter happy.
121 * Currently, we fill in the PMTU entry because netfilter
122 * refragmentation needs it, and the rt_flags entry because
123 * ipt_REJECT needs it. Future netfilter modules might
124 * require us to fill additional fields.
126 static const u32 br_dst_default_metrics
[RTAX_MAX
] = {
127 [RTAX_MTU
- 1] = 1500,
130 void br_netfilter_rtable_init(struct net_bridge
*br
)
132 struct rtable
*rt
= &br
->fake_rtable
;
134 atomic_set(&rt
->dst
.__refcnt
, 1);
135 rt
->dst
.dev
= br
->dev
;
136 rt
->dst
.path
= &rt
->dst
;
137 dst_init_metrics(&rt
->dst
, br_dst_default_metrics
, true);
138 rt
->dst
.flags
= DST_NOXFRM
;
139 rt
->dst
.ops
= &fake_dst_ops
;
142 static inline struct rtable
*bridge_parent_rtable(const struct net_device
*dev
)
144 struct net_bridge_port
*port
;
146 port
= br_port_get_rcu(dev
);
147 return port
? &port
->br
->fake_rtable
: NULL
;
150 static inline struct net_device
*bridge_parent(const struct net_device
*dev
)
152 struct net_bridge_port
*port
;
154 port
= br_port_get_rcu(dev
);
155 return port
? port
->br
->dev
: NULL
;
158 static inline struct nf_bridge_info
*nf_bridge_alloc(struct sk_buff
*skb
)
160 skb
->nf_bridge
= kzalloc(sizeof(struct nf_bridge_info
), GFP_ATOMIC
);
161 if (likely(skb
->nf_bridge
))
162 atomic_set(&(skb
->nf_bridge
->use
), 1);
164 return skb
->nf_bridge
;
167 static inline struct nf_bridge_info
*nf_bridge_unshare(struct sk_buff
*skb
)
169 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
171 if (atomic_read(&nf_bridge
->use
) > 1) {
172 struct nf_bridge_info
*tmp
= nf_bridge_alloc(skb
);
175 memcpy(tmp
, nf_bridge
, sizeof(struct nf_bridge_info
));
176 atomic_set(&tmp
->use
, 1);
178 nf_bridge_put(nf_bridge
);
184 static inline void nf_bridge_push_encap_header(struct sk_buff
*skb
)
186 unsigned int len
= nf_bridge_encap_header_len(skb
);
189 skb
->network_header
-= len
;
192 static inline void nf_bridge_pull_encap_header(struct sk_buff
*skb
)
194 unsigned int len
= nf_bridge_encap_header_len(skb
);
197 skb
->network_header
+= len
;
200 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff
*skb
)
202 unsigned int len
= nf_bridge_encap_header_len(skb
);
204 skb_pull_rcsum(skb
, len
);
205 skb
->network_header
+= len
;
208 static inline void nf_bridge_save_header(struct sk_buff
*skb
)
210 int header_size
= ETH_HLEN
+ nf_bridge_encap_header_len(skb
);
212 skb_copy_from_linear_data_offset(skb
, -header_size
,
213 skb
->nf_bridge
->data
, header_size
);
216 static inline void nf_bridge_update_protocol(struct sk_buff
*skb
)
218 if (skb
->nf_bridge
->mask
& BRNF_8021Q
)
219 skb
->protocol
= htons(ETH_P_8021Q
);
220 else if (skb
->nf_bridge
->mask
& BRNF_PPPoE
)
221 skb
->protocol
= htons(ETH_P_PPP_SES
);
224 /* When handing a packet over to the IP layer
225 * check whether we have a skb that is in the
229 static int br_parse_ip_options(struct sk_buff
*skb
)
231 struct ip_options
*opt
;
232 const struct iphdr
*iph
;
233 struct net_device
*dev
= skb
->dev
;
237 opt
= &(IPCB(skb
)->opt
);
239 /* Basic sanity checks */
240 if (iph
->ihl
< 5 || iph
->version
!= 4)
243 if (!pskb_may_pull(skb
, iph
->ihl
*4))
247 if (unlikely(ip_fast_csum((u8
*)iph
, iph
->ihl
)))
250 len
= ntohs(iph
->tot_len
);
251 if (skb
->len
< len
) {
252 IP_INC_STATS_BH(dev_net(dev
), IPSTATS_MIB_INTRUNCATEDPKTS
);
254 } else if (len
< (iph
->ihl
*4))
257 if (pskb_trim_rcsum(skb
, len
)) {
258 IP_INC_STATS_BH(dev_net(dev
), IPSTATS_MIB_INDISCARDS
);
262 memset(IPCB(skb
), 0, sizeof(struct inet_skb_parm
));
266 opt
->optlen
= iph
->ihl
*4 - sizeof(struct iphdr
);
267 if (ip_options_compile(dev_net(dev
), opt
, skb
))
270 /* Check correct handling of SRR option */
271 if (unlikely(opt
->srr
)) {
272 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
273 if (in_dev
&& !IN_DEV_SOURCE_ROUTE(in_dev
))
276 if (ip_options_rcv_srr(skb
))
283 IP_INC_STATS_BH(dev_net(dev
), IPSTATS_MIB_INHDRERRORS
);
288 /* Fill in the header for fragmented IP packets handled by
289 * the IPv4 connection tracking code.
291 int nf_bridge_copy_header(struct sk_buff
*skb
)
294 unsigned int header_size
;
296 nf_bridge_update_protocol(skb
);
297 header_size
= ETH_HLEN
+ nf_bridge_encap_header_len(skb
);
298 err
= skb_cow_head(skb
, header_size
);
302 skb_copy_to_linear_data_offset(skb
, -header_size
,
303 skb
->nf_bridge
->data
, header_size
);
304 __skb_push(skb
, nf_bridge_encap_header_len(skb
));
308 /* PF_BRIDGE/PRE_ROUTING *********************************************/
309 /* Undo the changes made for ip6tables PREROUTING and continue the
310 * bridge PRE_ROUTING hook. */
311 static int br_nf_pre_routing_finish_ipv6(struct sk_buff
*skb
)
313 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
316 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
317 skb
->pkt_type
= PACKET_OTHERHOST
;
318 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
320 nf_bridge
->mask
^= BRNF_NF_BRIDGE_PREROUTING
;
322 rt
= bridge_parent_rtable(nf_bridge
->physindev
);
327 skb_dst_set_noref(skb
, &rt
->dst
);
329 skb
->dev
= nf_bridge
->physindev
;
330 nf_bridge_update_protocol(skb
);
331 nf_bridge_push_encap_header(skb
);
332 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
333 br_handle_frame_finish
, 1);
338 /* Obtain the correct destination MAC address, while preserving the original
339 * source MAC address. If we already know this address, we just copy it. If we
340 * don't, we use the neighbour framework to find out. In both cases, we make
341 * sure that br_handle_frame_finish() is called afterwards.
343 static int br_nf_pre_routing_finish_bridge(struct sk_buff
*skb
)
345 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
346 struct dst_entry
*dst
;
348 skb
->dev
= bridge_parent(skb
->dev
);
353 neigh_hh_bridge(dst
->hh
, skb
);
354 skb
->dev
= nf_bridge
->physindev
;
355 return br_handle_frame_finish(skb
);
356 } else if (dst
->neighbour
) {
357 /* the neighbour function below overwrites the complete
358 * MAC header, so we save the Ethernet source address and
359 * protocol number. */
360 skb_copy_from_linear_data_offset(skb
, -(ETH_HLEN
-ETH_ALEN
), skb
->nf_bridge
->data
, ETH_HLEN
-ETH_ALEN
);
361 /* tell br_dev_xmit to continue with forwarding */
362 nf_bridge
->mask
|= BRNF_BRIDGED_DNAT
;
363 return dst
->neighbour
->output(skb
);
370 /* This requires some explaining. If DNAT has taken place,
371 * we will need to fix up the destination Ethernet address.
373 * There are two cases to consider:
374 * 1. The packet was DNAT'ed to a device in the same bridge
375 * port group as it was received on. We can still bridge
377 * 2. The packet was DNAT'ed to a different device, either
378 * a non-bridged device or another bridge port group.
379 * The packet will need to be routed.
381 * The correct way of distinguishing between these two cases is to
382 * call ip_route_input() and to look at skb->dst->dev, which is
383 * changed to the destination device if ip_route_input() succeeds.
385 * Let's first consider the case that ip_route_input() succeeds:
387 * If the output device equals the logical bridge device the packet
388 * came in on, we can consider this bridging. The corresponding MAC
389 * address will be obtained in br_nf_pre_routing_finish_bridge.
390 * Otherwise, the packet is considered to be routed and we just
391 * change the destination MAC address so that the packet will
392 * later be passed up to the IP stack to be routed. For a redirected
393 * packet, ip_route_input() will give back the localhost as output device,
394 * which differs from the bridge device.
396 * Let's now consider the case that ip_route_input() fails:
398 * This can be because the destination address is martian, in which case
399 * the packet will be dropped.
400 * If IP forwarding is disabled, ip_route_input() will fail, while
401 * ip_route_output_key() can return success. The source
402 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
403 * thinks we're handling a locally generated packet and won't care
404 * if IP forwarding is enabled. If the output device equals the logical bridge
405 * device, we proceed as if ip_route_input() succeeded. If it differs from the
406 * logical bridge port or if ip_route_output_key() fails we drop the packet.
408 static int br_nf_pre_routing_finish(struct sk_buff
*skb
)
410 struct net_device
*dev
= skb
->dev
;
411 struct iphdr
*iph
= ip_hdr(skb
);
412 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
416 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
417 skb
->pkt_type
= PACKET_OTHERHOST
;
418 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
420 nf_bridge
->mask
^= BRNF_NF_BRIDGE_PREROUTING
;
421 if (dnat_took_place(skb
)) {
422 if ((err
= ip_route_input(skb
, iph
->daddr
, iph
->saddr
, iph
->tos
, dev
))) {
423 struct in_device
*in_dev
= __in_dev_get_rcu(dev
);
425 /* If err equals -EHOSTUNREACH the error is due to a
426 * martian destination or due to the fact that
427 * forwarding is disabled. For most martian packets,
428 * ip_route_output_key() will fail. It won't fail for 2 types of
429 * martian destinations: loopback destinations and destination
430 * 0.0.0.0. In both cases the packet will be dropped because the
431 * destination is the loopback device and not the bridge. */
432 if (err
!= -EHOSTUNREACH
|| !in_dev
|| IN_DEV_FORWARD(in_dev
))
435 rt
= ip_route_output(dev_net(dev
), iph
->daddr
, 0,
436 RT_TOS(iph
->tos
), 0);
438 /* - Bridged-and-DNAT'ed traffic doesn't
439 * require ip_forwarding. */
440 if (rt
->dst
.dev
== dev
) {
441 skb_dst_set(skb
, &rt
->dst
);
450 if (skb_dst(skb
)->dev
== dev
) {
452 skb
->dev
= nf_bridge
->physindev
;
453 nf_bridge_update_protocol(skb
);
454 nf_bridge_push_encap_header(skb
);
455 NF_HOOK_THRESH(NFPROTO_BRIDGE
,
458 br_nf_pre_routing_finish_bridge
,
462 memcpy(eth_hdr(skb
)->h_dest
, dev
->dev_addr
, ETH_ALEN
);
463 skb
->pkt_type
= PACKET_HOST
;
466 rt
= bridge_parent_rtable(nf_bridge
->physindev
);
471 skb_dst_set_noref(skb
, &rt
->dst
);
474 skb
->dev
= nf_bridge
->physindev
;
475 nf_bridge_update_protocol(skb
);
476 nf_bridge_push_encap_header(skb
);
477 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
478 br_handle_frame_finish
, 1);
483 /* Some common code for IPv4/IPv6 */
484 static struct net_device
*setup_pre_routing(struct sk_buff
*skb
)
486 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
488 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
489 skb
->pkt_type
= PACKET_HOST
;
490 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
493 nf_bridge
->mask
|= BRNF_NF_BRIDGE_PREROUTING
;
494 nf_bridge
->physindev
= skb
->dev
;
495 skb
->dev
= bridge_parent(skb
->dev
);
496 if (skb
->protocol
== htons(ETH_P_8021Q
))
497 nf_bridge
->mask
|= BRNF_8021Q
;
498 else if (skb
->protocol
== htons(ETH_P_PPP_SES
))
499 nf_bridge
->mask
|= BRNF_PPPoE
;
504 /* We only check the length. A bridge shouldn't do any hop-by-hop stuff anyway */
505 static int check_hbh_len(struct sk_buff
*skb
)
507 unsigned char *raw
= (u8
*)(ipv6_hdr(skb
) + 1);
509 const unsigned char *nh
= skb_network_header(skb
);
511 int len
= (raw
[1] + 1) << 3;
513 if ((raw
+ len
) - skb
->data
> skb_headlen(skb
))
520 int optlen
= nh
[off
+ 1] + 2;
531 if (nh
[off
+ 1] != 4 || (off
& 3) != 2)
533 pkt_len
= ntohl(*(__be32
*) (nh
+ off
+ 2));
534 if (pkt_len
<= IPV6_MAXPLEN
||
535 ipv6_hdr(skb
)->payload_len
)
537 if (pkt_len
> skb
->len
- sizeof(struct ipv6hdr
))
539 if (pskb_trim_rcsum(skb
,
540 pkt_len
+ sizeof(struct ipv6hdr
)))
542 nh
= skb_network_header(skb
);
559 /* Replicate the checks that IPv6 does on packet reception and pass the packet
560 * to ip6tables, which doesn't support NAT, so things are fairly simple. */
561 static unsigned int br_nf_pre_routing_ipv6(unsigned int hook
,
563 const struct net_device
*in
,
564 const struct net_device
*out
,
565 int (*okfn
)(struct sk_buff
*))
567 const struct ipv6hdr
*hdr
;
570 if (skb
->len
< sizeof(struct ipv6hdr
))
573 if (!pskb_may_pull(skb
, sizeof(struct ipv6hdr
)))
578 if (hdr
->version
!= 6)
581 pkt_len
= ntohs(hdr
->payload_len
);
583 if (pkt_len
|| hdr
->nexthdr
!= NEXTHDR_HOP
) {
584 if (pkt_len
+ sizeof(struct ipv6hdr
) > skb
->len
)
586 if (pskb_trim_rcsum(skb
, pkt_len
+ sizeof(struct ipv6hdr
)))
589 if (hdr
->nexthdr
== NEXTHDR_HOP
&& check_hbh_len(skb
))
592 nf_bridge_put(skb
->nf_bridge
);
593 if (!nf_bridge_alloc(skb
))
595 if (!setup_pre_routing(skb
))
598 skb
->protocol
= htons(ETH_P_IPV6
);
599 NF_HOOK(NFPROTO_IPV6
, NF_INET_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
600 br_nf_pre_routing_finish_ipv6
);
605 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
606 * Replicate the checks that IPv4 does on packet reception.
607 * Set skb->dev to the bridge device (i.e. parent of the
608 * receiving device) to make netfilter happy, the REDIRECT
609 * target in particular. Save the original destination IP
610 * address to be able to detect DNAT afterwards. */
611 static unsigned int br_nf_pre_routing(unsigned int hook
, struct sk_buff
*skb
,
612 const struct net_device
*in
,
613 const struct net_device
*out
,
614 int (*okfn
)(struct sk_buff
*))
616 struct net_bridge_port
*p
;
617 struct net_bridge
*br
;
618 __u32 len
= nf_bridge_encap_header_len(skb
);
620 if (unlikely(!pskb_may_pull(skb
, len
)))
623 p
= br_port_get_rcu(in
);
628 if (skb
->protocol
== htons(ETH_P_IPV6
) || IS_VLAN_IPV6(skb
) ||
629 IS_PPPOE_IPV6(skb
)) {
630 if (!brnf_call_ip6tables
&& !br
->nf_call_ip6tables
)
633 nf_bridge_pull_encap_header_rcsum(skb
);
634 return br_nf_pre_routing_ipv6(hook
, skb
, in
, out
, okfn
);
637 if (!brnf_call_iptables
&& !br
->nf_call_iptables
)
640 if (skb
->protocol
!= htons(ETH_P_IP
) && !IS_VLAN_IP(skb
) &&
644 nf_bridge_pull_encap_header_rcsum(skb
);
646 if (br_parse_ip_options(skb
))
649 nf_bridge_put(skb
->nf_bridge
);
650 if (!nf_bridge_alloc(skb
))
652 if (!setup_pre_routing(skb
))
654 store_orig_dstaddr(skb
);
655 skb
->protocol
= htons(ETH_P_IP
);
657 NF_HOOK(NFPROTO_IPV4
, NF_INET_PRE_ROUTING
, skb
, skb
->dev
, NULL
,
658 br_nf_pre_routing_finish
);
664 /* PF_BRIDGE/LOCAL_IN ************************************************/
665 /* The packet is locally destined, which requires a real
666 * dst_entry, so detach the fake one. On the way up, the
667 * packet would pass through PRE_ROUTING again (which already
668 * took place when the packet entered the bridge), but we
669 * register an IPv4 PRE_ROUTING 'sabotage' hook that will
670 * prevent this from happening. */
671 static unsigned int br_nf_local_in(unsigned int hook
, struct sk_buff
*skb
,
672 const struct net_device
*in
,
673 const struct net_device
*out
,
674 int (*okfn
)(struct sk_buff
*))
676 struct rtable
*rt
= skb_rtable(skb
);
678 if (rt
&& rt
== bridge_parent_rtable(in
))
684 /* PF_BRIDGE/FORWARD *************************************************/
685 static int br_nf_forward_finish(struct sk_buff
*skb
)
687 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
688 struct net_device
*in
;
690 if (skb
->protocol
!= htons(ETH_P_ARP
) && !IS_VLAN_ARP(skb
)) {
691 in
= nf_bridge
->physindev
;
692 if (nf_bridge
->mask
& BRNF_PKT_TYPE
) {
693 skb
->pkt_type
= PACKET_OTHERHOST
;
694 nf_bridge
->mask
^= BRNF_PKT_TYPE
;
696 nf_bridge_update_protocol(skb
);
698 in
= *((struct net_device
**)(skb
->cb
));
700 nf_bridge_push_encap_header(skb
);
702 NF_HOOK_THRESH(NFPROTO_BRIDGE
, NF_BR_FORWARD
, skb
, in
,
703 skb
->dev
, br_forward_finish
, 1);
707 /* This is the 'purely bridged' case. For IP, we pass the packet to
708 * netfilter with indev and outdev set to the bridge device,
709 * but we are still able to filter on the 'real' indev/outdev
710 * because of the physdev module. For ARP, indev and outdev are the
712 static unsigned int br_nf_forward_ip(unsigned int hook
, struct sk_buff
*skb
,
713 const struct net_device
*in
,
714 const struct net_device
*out
,
715 int (*okfn
)(struct sk_buff
*))
717 struct nf_bridge_info
*nf_bridge
;
718 struct net_device
*parent
;
724 /* Need exclusive nf_bridge_info since we might have multiple
725 * different physoutdevs. */
726 if (!nf_bridge_unshare(skb
))
729 parent
= bridge_parent(out
);
733 if (skb
->protocol
== htons(ETH_P_IP
) || IS_VLAN_IP(skb
) ||
736 else if (skb
->protocol
== htons(ETH_P_IPV6
) || IS_VLAN_IPV6(skb
) ||
742 nf_bridge_pull_encap_header(skb
);
744 nf_bridge
= skb
->nf_bridge
;
745 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
746 skb
->pkt_type
= PACKET_HOST
;
747 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
750 if (pf
== PF_INET
&& br_parse_ip_options(skb
))
753 /* The physdev module checks on this */
754 nf_bridge
->mask
|= BRNF_BRIDGED
;
755 nf_bridge
->physoutdev
= skb
->dev
;
757 skb
->protocol
= htons(ETH_P_IP
);
759 skb
->protocol
= htons(ETH_P_IPV6
);
761 NF_HOOK(pf
, NF_INET_FORWARD
, skb
, bridge_parent(in
), parent
,
762 br_nf_forward_finish
);
767 static unsigned int br_nf_forward_arp(unsigned int hook
, struct sk_buff
*skb
,
768 const struct net_device
*in
,
769 const struct net_device
*out
,
770 int (*okfn
)(struct sk_buff
*))
772 struct net_bridge_port
*p
;
773 struct net_bridge
*br
;
774 struct net_device
**d
= (struct net_device
**)(skb
->cb
);
776 p
= br_port_get_rcu(out
);
781 if (!brnf_call_arptables
&& !br
->nf_call_arptables
)
784 if (skb
->protocol
!= htons(ETH_P_ARP
)) {
785 if (!IS_VLAN_ARP(skb
))
787 nf_bridge_pull_encap_header(skb
);
790 if (arp_hdr(skb
)->ar_pln
!= 4) {
791 if (IS_VLAN_ARP(skb
))
792 nf_bridge_push_encap_header(skb
);
795 *d
= (struct net_device
*)in
;
796 NF_HOOK(NFPROTO_ARP
, NF_ARP_FORWARD
, skb
, (struct net_device
*)in
,
797 (struct net_device
*)out
, br_nf_forward_finish
);
802 #if defined(CONFIG_NF_CONNTRACK_IPV4) || defined(CONFIG_NF_CONNTRACK_IPV4_MODULE)
803 static int br_nf_dev_queue_xmit(struct sk_buff
*skb
)
807 if (skb
->nfct
!= NULL
&& skb
->protocol
== htons(ETH_P_IP
) &&
808 skb
->len
+ nf_bridge_mtu_reduction(skb
) > skb
->dev
->mtu
&&
810 if (br_parse_ip_options(skb
))
811 /* Drop invalid packet */
813 ret
= ip_fragment(skb
, br_dev_queue_push_xmit
);
815 ret
= br_dev_queue_push_xmit(skb
);
820 static int br_nf_dev_queue_xmit(struct sk_buff
*skb
)
822 return br_dev_queue_push_xmit(skb
);
826 /* PF_BRIDGE/POST_ROUTING ********************************************/
827 static unsigned int br_nf_post_routing(unsigned int hook
, struct sk_buff
*skb
,
828 const struct net_device
*in
,
829 const struct net_device
*out
,
830 int (*okfn
)(struct sk_buff
*))
832 struct nf_bridge_info
*nf_bridge
= skb
->nf_bridge
;
833 struct net_device
*realoutdev
= bridge_parent(skb
->dev
);
836 if (!nf_bridge
|| !(nf_bridge
->mask
& BRNF_BRIDGED
))
842 if (skb
->protocol
== htons(ETH_P_IP
) || IS_VLAN_IP(skb
) ||
845 else if (skb
->protocol
== htons(ETH_P_IPV6
) || IS_VLAN_IPV6(skb
) ||
851 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
852 * about the value of skb->pkt_type. */
853 if (skb
->pkt_type
== PACKET_OTHERHOST
) {
854 skb
->pkt_type
= PACKET_HOST
;
855 nf_bridge
->mask
|= BRNF_PKT_TYPE
;
858 nf_bridge_pull_encap_header(skb
);
859 nf_bridge_save_header(skb
);
861 skb
->protocol
= htons(ETH_P_IP
);
863 skb
->protocol
= htons(ETH_P_IPV6
);
865 NF_HOOK(pf
, NF_INET_POST_ROUTING
, skb
, NULL
, realoutdev
,
866 br_nf_dev_queue_xmit
);
871 /* IP/SABOTAGE *****************************************************/
872 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
873 * for the second time. */
874 static unsigned int ip_sabotage_in(unsigned int hook
, struct sk_buff
*skb
,
875 const struct net_device
*in
,
876 const struct net_device
*out
,
877 int (*okfn
)(struct sk_buff
*))
879 if (skb
->nf_bridge
&&
880 !(skb
->nf_bridge
->mask
& BRNF_NF_BRIDGE_PREROUTING
)) {
887 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
888 * br_dev_queue_push_xmit is called afterwards */
889 static struct nf_hook_ops br_nf_ops
[] __read_mostly
= {
891 .hook
= br_nf_pre_routing
,
892 .owner
= THIS_MODULE
,
894 .hooknum
= NF_BR_PRE_ROUTING
,
895 .priority
= NF_BR_PRI_BRNF
,
898 .hook
= br_nf_local_in
,
899 .owner
= THIS_MODULE
,
901 .hooknum
= NF_BR_LOCAL_IN
,
902 .priority
= NF_BR_PRI_BRNF
,
905 .hook
= br_nf_forward_ip
,
906 .owner
= THIS_MODULE
,
908 .hooknum
= NF_BR_FORWARD
,
909 .priority
= NF_BR_PRI_BRNF
- 1,
912 .hook
= br_nf_forward_arp
,
913 .owner
= THIS_MODULE
,
915 .hooknum
= NF_BR_FORWARD
,
916 .priority
= NF_BR_PRI_BRNF
,
919 .hook
= br_nf_post_routing
,
920 .owner
= THIS_MODULE
,
922 .hooknum
= NF_BR_POST_ROUTING
,
923 .priority
= NF_BR_PRI_LAST
,
926 .hook
= ip_sabotage_in
,
927 .owner
= THIS_MODULE
,
929 .hooknum
= NF_INET_PRE_ROUTING
,
930 .priority
= NF_IP_PRI_FIRST
,
933 .hook
= ip_sabotage_in
,
934 .owner
= THIS_MODULE
,
936 .hooknum
= NF_INET_PRE_ROUTING
,
937 .priority
= NF_IP6_PRI_FIRST
,
943 int brnf_sysctl_call_tables(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 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
,
994 static struct ctl_path brnf_path
[] = {
995 { .procname
= "net", },
996 { .procname
= "bridge", },
1001 int __init
br_netfilter_init(void)
1005 ret
= dst_entries_init(&fake_dst_ops
);
1009 ret
= nf_register_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
1011 dst_entries_destroy(&fake_dst_ops
);
1014 #ifdef CONFIG_SYSCTL
1015 brnf_sysctl_header
= register_sysctl_paths(brnf_path
, brnf_table
);
1016 if (brnf_sysctl_header
== NULL
) {
1018 "br_netfilter: can't register to sysctl.\n");
1019 nf_unregister_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
1020 dst_entries_destroy(&fake_dst_ops
);
1024 printk(KERN_NOTICE
"Bridge firewalling registered\n");
1028 void br_netfilter_fini(void)
1030 nf_unregister_hooks(br_nf_ops
, ARRAY_SIZE(br_nf_ops
));
1031 #ifdef CONFIG_SYSCTL
1032 unregister_sysctl_table(brnf_sysctl_header
);
1034 dst_entries_destroy(&fake_dst_ops
);