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 static inline struct rtable *bridge_parent_rtable(const struct net_device *dev)
116 struct net_bridge_port *port;
118 port = br_port_get_rcu(dev);
119 return port ? &port->br->fake_rtable : NULL;
122 static inline struct net_device *bridge_parent(const struct net_device *dev)
124 struct net_bridge_port *port;
126 port = br_port_get_rcu(dev);
127 return port ? port->br->dev : NULL;
130 static inline struct nf_bridge_info *nf_bridge_alloc(struct sk_buff *skb)
132 skb->nf_bridge = kzalloc(sizeof(struct nf_bridge_info), GFP_ATOMIC);
133 if (likely(skb->nf_bridge))
134 atomic_set(&(skb->nf_bridge->use), 1);
136 return skb->nf_bridge;
139 static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
141 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
143 if (atomic_read(&nf_bridge->use) > 1) {
144 struct nf_bridge_info *tmp = nf_bridge_alloc(skb);
147 memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info));
148 atomic_set(&tmp->use, 1);
150 nf_bridge_put(nf_bridge);
156 static inline void nf_bridge_push_encap_header(struct sk_buff *skb)
158 unsigned int len = nf_bridge_encap_header_len(skb);
161 skb->network_header -= len;
164 static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
166 unsigned int len = nf_bridge_encap_header_len(skb);
169 skb->network_header += len;
172 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
174 unsigned int len = nf_bridge_encap_header_len(skb);
176 skb_pull_rcsum(skb, len);
177 skb->network_header += len;
180 static inline void nf_bridge_save_header(struct sk_buff *skb)
182 int header_size = ETH_HLEN + nf_bridge_encap_header_len(skb);
184 skb_copy_from_linear_data_offset(skb, -header_size,
185 skb->nf_bridge->data, header_size);
188 /* When handing a packet over to the IP layer
189 * check whether we have a skb that is in the
193 static int br_parse_ip_options(struct sk_buff *skb)
195 const struct iphdr *iph;
196 struct net_device *dev = skb->dev;
199 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
204 /* Basic sanity checks */
205 if (iph->ihl < 5 || iph->version != 4)
208 if (!pskb_may_pull(skb, iph->ihl*4))
212 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
215 len = ntohs(iph->tot_len);
216 if (skb->len < len) {
217 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INTRUNCATEDPKTS);
219 } else if (len < (iph->ihl*4))
222 if (pskb_trim_rcsum(skb, len)) {
223 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
227 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
228 /* We should really parse IP options here but until
229 * somebody who actually uses IP options complains to
230 * us we'll just silently ignore the options because
236 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
241 static void nf_bridge_update_protocol(struct sk_buff *skb)
243 if (skb->nf_bridge->mask & BRNF_8021Q)
244 skb->protocol = htons(ETH_P_8021Q);
245 else if (skb->nf_bridge->mask & BRNF_PPPoE)
246 skb->protocol = htons(ETH_P_PPP_SES);
249 /* PF_BRIDGE/PRE_ROUTING *********************************************/
250 /* Undo the changes made for ip6tables PREROUTING and continue the
251 * bridge PRE_ROUTING hook. */
252 static int br_nf_pre_routing_finish_ipv6(struct sk_buff *skb)
254 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
257 if (nf_bridge->mask & BRNF_PKT_TYPE) {
258 skb->pkt_type = PACKET_OTHERHOST;
259 nf_bridge->mask ^= BRNF_PKT_TYPE;
261 nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
263 rt = bridge_parent_rtable(nf_bridge->physindev);
268 skb_dst_set_noref(skb, &rt->dst);
270 skb->dev = nf_bridge->physindev;
271 nf_bridge_update_protocol(skb);
272 nf_bridge_push_encap_header(skb);
273 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
274 br_handle_frame_finish, 1);
279 /* Obtain the correct destination MAC address, while preserving the original
280 * source MAC address. If we already know this address, we just copy it. If we
281 * don't, we use the neighbour framework to find out. In both cases, we make
282 * sure that br_handle_frame_finish() is called afterwards.
284 static int br_nf_pre_routing_finish_bridge(struct sk_buff *skb)
286 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
287 struct neighbour *neigh;
288 struct dst_entry *dst;
290 skb->dev = bridge_parent(skb->dev);
294 neigh = dst_neigh_lookup_skb(dst, skb);
298 if (neigh->hh.hh_len) {
299 neigh_hh_bridge(&neigh->hh, skb);
300 skb->dev = nf_bridge->physindev;
301 ret = br_handle_frame_finish(skb);
303 /* the neighbour function below overwrites the complete
304 * MAC header, so we save the Ethernet source address and
307 skb_copy_from_linear_data_offset(skb,
308 -(ETH_HLEN-ETH_ALEN),
309 skb->nf_bridge->data,
311 /* tell br_dev_xmit to continue with forwarding */
312 nf_bridge->mask |= BRNF_BRIDGED_DNAT;
313 /* FIXME Need to refragment */
314 ret = neigh->output(neigh, skb);
316 neigh_release(neigh);
324 static bool dnat_took_place(const struct sk_buff *skb)
326 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
327 enum ip_conntrack_info ctinfo;
330 ct = nf_ct_get(skb, &ctinfo);
331 if (!ct || nf_ct_is_untracked(ct))
334 return test_bit(IPS_DST_NAT_BIT, &ct->status);
340 /* This requires some explaining. If DNAT has taken place,
341 * we will need to fix up the destination Ethernet address.
343 * There are two cases to consider:
344 * 1. The packet was DNAT'ed to a device in the same bridge
345 * port group as it was received on. We can still bridge
347 * 2. The packet was DNAT'ed to a different device, either
348 * a non-bridged device or another bridge port group.
349 * The packet will need to be routed.
351 * The correct way of distinguishing between these two cases is to
352 * call ip_route_input() and to look at skb->dst->dev, which is
353 * changed to the destination device if ip_route_input() succeeds.
355 * Let's first consider the case that ip_route_input() succeeds:
357 * If the output device equals the logical bridge device the packet
358 * came in on, we can consider this bridging. The corresponding MAC
359 * address will be obtained in br_nf_pre_routing_finish_bridge.
360 * Otherwise, the packet is considered to be routed and we just
361 * change the destination MAC address so that the packet will
362 * later be passed up to the IP stack to be routed. For a redirected
363 * packet, ip_route_input() will give back the localhost as output device,
364 * which differs from the bridge device.
366 * Let's now consider the case that ip_route_input() fails:
368 * This can be because the destination address is martian, in which case
369 * the packet will be dropped.
370 * If IP forwarding is disabled, ip_route_input() will fail, while
371 * ip_route_output_key() can return success. The source
372 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
373 * thinks we're handling a locally generated packet and won't care
374 * if IP forwarding is enabled. If the output device equals the logical bridge
375 * device, we proceed as if ip_route_input() succeeded. If it differs from the
376 * logical bridge port or if ip_route_output_key() fails we drop the packet.
378 static int br_nf_pre_routing_finish(struct sk_buff *skb)
380 struct net_device *dev = skb->dev;
381 struct iphdr *iph = ip_hdr(skb);
382 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
387 frag_max_size = IPCB(skb)->frag_max_size;
388 BR_INPUT_SKB_CB(skb)->frag_max_size = frag_max_size;
390 if (nf_bridge->mask & BRNF_PKT_TYPE) {
391 skb->pkt_type = PACKET_OTHERHOST;
392 nf_bridge->mask ^= BRNF_PKT_TYPE;
394 nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
395 if (dnat_took_place(skb)) {
396 if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
397 struct in_device *in_dev = __in_dev_get_rcu(dev);
399 /* If err equals -EHOSTUNREACH the error is due to a
400 * martian destination or due to the fact that
401 * forwarding is disabled. For most martian packets,
402 * ip_route_output_key() will fail. It won't fail for 2 types of
403 * martian destinations: loopback destinations and destination
404 * 0.0.0.0. In both cases the packet will be dropped because the
405 * destination is the loopback device and not the bridge. */
406 if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
409 rt = ip_route_output(dev_net(dev), iph->daddr, 0,
410 RT_TOS(iph->tos), 0);
412 /* - Bridged-and-DNAT'ed traffic doesn't
413 * require ip_forwarding. */
414 if (rt->dst.dev == dev) {
415 skb_dst_set(skb, &rt->dst);
424 if (skb_dst(skb)->dev == dev) {
426 skb->dev = nf_bridge->physindev;
427 nf_bridge_update_protocol(skb);
428 nf_bridge_push_encap_header(skb);
429 NF_HOOK_THRESH(NFPROTO_BRIDGE,
432 br_nf_pre_routing_finish_bridge,
436 ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr);
437 skb->pkt_type = PACKET_HOST;
440 rt = bridge_parent_rtable(nf_bridge->physindev);
445 skb_dst_set_noref(skb, &rt->dst);
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, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
452 br_handle_frame_finish, 1);
457 static struct net_device *brnf_get_logical_dev(struct sk_buff *skb, const struct net_device *dev)
459 struct net_device *vlan, *br;
461 br = bridge_parent(dev);
462 if (brnf_pass_vlan_indev == 0 || !skb_vlan_tag_present(skb))
465 vlan = __vlan_find_dev_deep_rcu(br, skb->vlan_proto,
466 skb_vlan_tag_get(skb) & VLAN_VID_MASK);
468 return vlan ? vlan : br;
471 /* Some common code for IPv4/IPv6 */
472 static struct net_device *setup_pre_routing(struct sk_buff *skb)
474 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
476 if (skb->pkt_type == PACKET_OTHERHOST) {
477 skb->pkt_type = PACKET_HOST;
478 nf_bridge->mask |= BRNF_PKT_TYPE;
481 nf_bridge->mask |= BRNF_NF_BRIDGE_PREROUTING;
482 nf_bridge->physindev = skb->dev;
483 skb->dev = brnf_get_logical_dev(skb, skb->dev);
484 if (skb->protocol == htons(ETH_P_8021Q))
485 nf_bridge->mask |= BRNF_8021Q;
486 else if (skb->protocol == htons(ETH_P_PPP_SES))
487 nf_bridge->mask |= BRNF_PPPoE;
489 /* Must drop socket now because of tproxy. */
494 /* We only check the length. A bridge shouldn't do any hop-by-hop stuff anyway */
495 static int check_hbh_len(struct sk_buff *skb)
497 unsigned char *raw = (u8 *)(ipv6_hdr(skb) + 1);
499 const unsigned char *nh = skb_network_header(skb);
501 int len = (raw[1] + 1) << 3;
503 if ((raw + len) - skb->data > skb_headlen(skb))
510 int optlen = nh[off + 1] + 2;
521 if (nh[off + 1] != 4 || (off & 3) != 2)
523 pkt_len = ntohl(*(__be32 *) (nh + off + 2));
524 if (pkt_len <= IPV6_MAXPLEN ||
525 ipv6_hdr(skb)->payload_len)
527 if (pkt_len > skb->len - sizeof(struct ipv6hdr))
529 if (pskb_trim_rcsum(skb,
530 pkt_len + sizeof(struct ipv6hdr)))
532 nh = skb_network_header(skb);
549 /* Replicate the checks that IPv6 does on packet reception and pass the packet
550 * to ip6tables, which doesn't support NAT, so things are fairly simple. */
551 static unsigned int br_nf_pre_routing_ipv6(const struct nf_hook_ops *ops,
553 const struct net_device *in,
554 const struct net_device *out,
555 int (*okfn)(struct sk_buff *))
557 const struct ipv6hdr *hdr;
560 if (skb->len < sizeof(struct ipv6hdr))
563 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
568 if (hdr->version != 6)
571 pkt_len = ntohs(hdr->payload_len);
573 if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
574 if (pkt_len + sizeof(struct ipv6hdr) > skb->len)
576 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
579 if (hdr->nexthdr == NEXTHDR_HOP && check_hbh_len(skb))
582 nf_bridge_put(skb->nf_bridge);
583 if (!nf_bridge_alloc(skb))
585 if (!setup_pre_routing(skb))
588 skb->protocol = htons(ETH_P_IPV6);
589 NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
590 br_nf_pre_routing_finish_ipv6);
595 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
596 * Replicate the checks that IPv4 does on packet reception.
597 * Set skb->dev to the bridge device (i.e. parent of the
598 * receiving device) to make netfilter happy, the REDIRECT
599 * target in particular. Save the original destination IP
600 * address to be able to detect DNAT afterwards. */
601 static unsigned int br_nf_pre_routing(const struct nf_hook_ops *ops,
603 const struct net_device *in,
604 const struct net_device *out,
605 int (*okfn)(struct sk_buff *))
607 struct net_bridge_port *p;
608 struct net_bridge *br;
609 __u32 len = nf_bridge_encap_header_len(skb);
611 if (unlikely(!pskb_may_pull(skb, len)))
614 p = br_port_get_rcu(in);
619 if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb)) {
620 if (!brnf_call_ip6tables && !br->nf_call_ip6tables)
623 nf_bridge_pull_encap_header_rcsum(skb);
624 return br_nf_pre_routing_ipv6(ops, skb, in, out, okfn);
627 if (!brnf_call_iptables && !br->nf_call_iptables)
630 if (!IS_IP(skb) && !IS_VLAN_IP(skb) && !IS_PPPOE_IP(skb))
633 nf_bridge_pull_encap_header_rcsum(skb);
635 if (br_parse_ip_options(skb))
638 nf_bridge_put(skb->nf_bridge);
639 if (!nf_bridge_alloc(skb))
641 if (!setup_pre_routing(skb))
644 skb->protocol = htons(ETH_P_IP);
646 NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
647 br_nf_pre_routing_finish);
653 /* PF_BRIDGE/LOCAL_IN ************************************************/
654 /* The packet is locally destined, which requires a real
655 * dst_entry, so detach the fake one. On the way up, the
656 * packet would pass through PRE_ROUTING again (which already
657 * took place when the packet entered the bridge), but we
658 * register an IPv4 PRE_ROUTING 'sabotage' hook that will
659 * prevent this from happening. */
660 static unsigned int br_nf_local_in(const struct nf_hook_ops *ops,
662 const struct net_device *in,
663 const struct net_device *out,
664 int (*okfn)(struct sk_buff *))
666 br_drop_fake_rtable(skb);
670 /* PF_BRIDGE/FORWARD *************************************************/
671 static int br_nf_forward_finish(struct sk_buff *skb)
673 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
674 struct net_device *in;
676 if (!IS_ARP(skb) && !IS_VLAN_ARP(skb)) {
677 in = nf_bridge->physindev;
678 if (nf_bridge->mask & BRNF_PKT_TYPE) {
679 skb->pkt_type = PACKET_OTHERHOST;
680 nf_bridge->mask ^= BRNF_PKT_TYPE;
682 nf_bridge_update_protocol(skb);
684 in = *((struct net_device **)(skb->cb));
686 nf_bridge_push_encap_header(skb);
688 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_FORWARD, skb, in,
689 skb->dev, br_forward_finish, 1);
694 /* This is the 'purely bridged' case. For IP, we pass the packet to
695 * netfilter with indev and outdev set to the bridge device,
696 * but we are still able to filter on the 'real' indev/outdev
697 * because of the physdev module. For ARP, indev and outdev are the
699 static unsigned int br_nf_forward_ip(const struct nf_hook_ops *ops,
701 const struct net_device *in,
702 const struct net_device *out,
703 int (*okfn)(struct sk_buff *))
705 struct nf_bridge_info *nf_bridge;
706 struct net_device *parent;
712 /* Need exclusive nf_bridge_info since we might have multiple
713 * different physoutdevs. */
714 if (!nf_bridge_unshare(skb))
717 parent = bridge_parent(out);
721 if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
723 else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
728 nf_bridge_pull_encap_header(skb);
730 nf_bridge = skb->nf_bridge;
731 if (skb->pkt_type == PACKET_OTHERHOST) {
732 skb->pkt_type = PACKET_HOST;
733 nf_bridge->mask |= BRNF_PKT_TYPE;
736 if (pf == NFPROTO_IPV4 && br_parse_ip_options(skb))
739 nf_bridge->physoutdev = skb->dev;
740 if (pf == NFPROTO_IPV4)
741 skb->protocol = htons(ETH_P_IP);
743 skb->protocol = htons(ETH_P_IPV6);
745 NF_HOOK(pf, NF_INET_FORWARD, skb, brnf_get_logical_dev(skb, in), parent,
746 br_nf_forward_finish);
751 static unsigned int br_nf_forward_arp(const struct nf_hook_ops *ops,
753 const struct net_device *in,
754 const struct net_device *out,
755 int (*okfn)(struct sk_buff *))
757 struct net_bridge_port *p;
758 struct net_bridge *br;
759 struct net_device **d = (struct net_device **)(skb->cb);
761 p = br_port_get_rcu(out);
766 if (!brnf_call_arptables && !br->nf_call_arptables)
770 if (!IS_VLAN_ARP(skb))
772 nf_bridge_pull_encap_header(skb);
775 if (arp_hdr(skb)->ar_pln != 4) {
776 if (IS_VLAN_ARP(skb))
777 nf_bridge_push_encap_header(skb);
780 *d = (struct net_device *)in;
781 NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, skb, (struct net_device *)in,
782 (struct net_device *)out, br_nf_forward_finish);
787 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4)
788 static bool nf_bridge_copy_header(struct sk_buff *skb)
791 unsigned int header_size;
793 nf_bridge_update_protocol(skb);
794 header_size = ETH_HLEN + nf_bridge_encap_header_len(skb);
795 err = skb_cow_head(skb, header_size);
799 skb_copy_to_linear_data_offset(skb, -header_size,
800 skb->nf_bridge->data, header_size);
801 __skb_push(skb, nf_bridge_encap_header_len(skb));
805 static int br_nf_push_frag_xmit(struct sk_buff *skb)
807 if (!nf_bridge_copy_header(skb)) {
812 return br_dev_queue_push_xmit(skb);
815 static int br_nf_dev_queue_xmit(struct sk_buff *skb)
819 unsigned int mtu_reserved;
821 if (skb_is_gso(skb) || skb->protocol != htons(ETH_P_IP))
822 return br_dev_queue_push_xmit(skb);
824 mtu_reserved = nf_bridge_mtu_reduction(skb);
825 /* This is wrong! We should preserve the original fragment
826 * boundaries by preserving frag_list rather than refragmenting.
828 if (skb->len + mtu_reserved > skb->dev->mtu) {
829 frag_max_size = BR_INPUT_SKB_CB(skb)->frag_max_size;
830 if (br_parse_ip_options(skb))
831 /* Drop invalid packet */
833 IPCB(skb)->frag_max_size = frag_max_size;
834 ret = ip_fragment(skb, br_nf_push_frag_xmit);
836 ret = br_dev_queue_push_xmit(skb);
841 static int br_nf_dev_queue_xmit(struct sk_buff *skb)
843 return br_dev_queue_push_xmit(skb);
847 /* PF_BRIDGE/POST_ROUTING ********************************************/
848 static unsigned int br_nf_post_routing(const struct nf_hook_ops *ops,
850 const struct net_device *in,
851 const struct net_device *out,
852 int (*okfn)(struct sk_buff *))
854 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
855 struct net_device *realoutdev = bridge_parent(skb->dev);
858 /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
859 * on a bridge, but was delivered locally and is now being routed:
861 * POST_ROUTING was already invoked from the ip stack.
863 if (!nf_bridge || !nf_bridge->physoutdev)
869 if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
871 else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
876 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
877 * about the value of skb->pkt_type. */
878 if (skb->pkt_type == PACKET_OTHERHOST) {
879 skb->pkt_type = PACKET_HOST;
880 nf_bridge->mask |= BRNF_PKT_TYPE;
883 nf_bridge_pull_encap_header(skb);
884 nf_bridge_save_header(skb);
885 if (pf == NFPROTO_IPV4)
886 skb->protocol = htons(ETH_P_IP);
888 skb->protocol = htons(ETH_P_IPV6);
890 NF_HOOK(pf, NF_INET_POST_ROUTING, skb, NULL, realoutdev,
891 br_nf_dev_queue_xmit);
896 /* IP/SABOTAGE *****************************************************/
897 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
898 * for the second time. */
899 static unsigned int ip_sabotage_in(const struct nf_hook_ops *ops,
901 const struct net_device *in,
902 const struct net_device *out,
903 int (*okfn)(struct sk_buff *))
905 if (skb->nf_bridge &&
906 !(skb->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)) {
913 /* This is called when br_netfilter has called into iptables/netfilter,
914 * and DNAT has taken place on a bridge-forwarded packet.
916 * neigh->output has created a new MAC header, with local br0 MAC
919 * This restores the original MAC saddr of the bridged packet
920 * before invoking bridge forward logic to transmit the packet.
922 static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff *skb)
924 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
926 skb_pull(skb, ETH_HLEN);
927 nf_bridge->mask &= ~BRNF_BRIDGED_DNAT;
929 skb_copy_to_linear_data_offset(skb, -(ETH_HLEN-ETH_ALEN),
930 skb->nf_bridge->data, ETH_HLEN-ETH_ALEN);
931 skb->dev = nf_bridge->physindev;
932 br_handle_frame_finish(skb);
935 static int br_nf_dev_xmit(struct sk_buff *skb)
937 if (skb->nf_bridge && (skb->nf_bridge->mask & BRNF_BRIDGED_DNAT)) {
938 br_nf_pre_routing_finish_bridge_slow(skb);
944 static const struct nf_br_ops br_ops = {
945 .br_dev_xmit_hook = br_nf_dev_xmit,
948 void br_netfilter_enable(void)
951 EXPORT_SYMBOL_GPL(br_netfilter_enable);
953 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
954 * br_dev_queue_push_xmit is called afterwards */
955 static struct nf_hook_ops br_nf_ops[] __read_mostly = {
957 .hook = br_nf_pre_routing,
958 .owner = THIS_MODULE,
959 .pf = NFPROTO_BRIDGE,
960 .hooknum = NF_BR_PRE_ROUTING,
961 .priority = NF_BR_PRI_BRNF,
964 .hook = br_nf_local_in,
965 .owner = THIS_MODULE,
966 .pf = NFPROTO_BRIDGE,
967 .hooknum = NF_BR_LOCAL_IN,
968 .priority = NF_BR_PRI_BRNF,
971 .hook = br_nf_forward_ip,
972 .owner = THIS_MODULE,
973 .pf = NFPROTO_BRIDGE,
974 .hooknum = NF_BR_FORWARD,
975 .priority = NF_BR_PRI_BRNF - 1,
978 .hook = br_nf_forward_arp,
979 .owner = THIS_MODULE,
980 .pf = NFPROTO_BRIDGE,
981 .hooknum = NF_BR_FORWARD,
982 .priority = NF_BR_PRI_BRNF,
985 .hook = br_nf_post_routing,
986 .owner = THIS_MODULE,
987 .pf = NFPROTO_BRIDGE,
988 .hooknum = NF_BR_POST_ROUTING,
989 .priority = NF_BR_PRI_LAST,
992 .hook = ip_sabotage_in,
993 .owner = THIS_MODULE,
995 .hooknum = NF_INET_PRE_ROUTING,
996 .priority = NF_IP_PRI_FIRST,
999 .hook = ip_sabotage_in,
1000 .owner = THIS_MODULE,
1002 .hooknum = NF_INET_PRE_ROUTING,
1003 .priority = NF_IP6_PRI_FIRST,
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 = nf_register_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
1076 #ifdef CONFIG_SYSCTL
1077 brnf_sysctl_header = register_net_sysctl(&init_net, "net/bridge", brnf_table);
1078 if (brnf_sysctl_header == NULL) {
1080 "br_netfilter: can't register to sysctl.\n");
1081 nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
1085 RCU_INIT_POINTER(nf_br_ops, &br_ops);
1086 printk(KERN_NOTICE "Bridge firewalling registered\n");
1090 static void __exit br_netfilter_fini(void)
1092 RCU_INIT_POINTER(nf_br_ops, NULL);
1093 nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
1094 #ifdef CONFIG_SYSCTL
1095 unregister_net_sysctl_table(brnf_sysctl_header);
1099 module_init(br_netfilter_init);
1100 module_exit(br_netfilter_fini);
1102 MODULE_LICENSE("GPL");
1103 MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
1104 MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
1105 MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");