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 #include <asm/uaccess.h>
41 #include "br_private.h"
43 #include <linux/sysctl.h>
47 static struct ctl_table_header *brnf_sysctl_header;
48 static int brnf_call_iptables __read_mostly = 1;
49 static int brnf_call_ip6tables __read_mostly = 1;
50 static int brnf_call_arptables __read_mostly = 1;
51 static int brnf_filter_vlan_tagged __read_mostly = 0;
52 static int brnf_filter_pppoe_tagged __read_mostly = 0;
53 static int brnf_pass_vlan_indev __read_mostly = 0;
55 #define brnf_call_iptables 1
56 #define brnf_call_ip6tables 1
57 #define brnf_call_arptables 1
58 #define brnf_filter_vlan_tagged 0
59 #define brnf_filter_pppoe_tagged 0
60 #define brnf_pass_vlan_indev 0
64 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP))
66 #define IS_IPV6(skb) \
67 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6))
70 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP))
72 static inline __be16 vlan_proto(const struct sk_buff *skb)
74 if (skb_vlan_tag_present(skb))
76 else if (skb->protocol == htons(ETH_P_8021Q))
77 return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
82 #define IS_VLAN_IP(skb) \
83 (vlan_proto(skb) == htons(ETH_P_IP) && \
84 brnf_filter_vlan_tagged)
86 #define IS_VLAN_IPV6(skb) \
87 (vlan_proto(skb) == htons(ETH_P_IPV6) && \
88 brnf_filter_vlan_tagged)
90 #define IS_VLAN_ARP(skb) \
91 (vlan_proto(skb) == htons(ETH_P_ARP) && \
92 brnf_filter_vlan_tagged)
94 static inline __be16 pppoe_proto(const struct sk_buff *skb)
96 return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
97 sizeof(struct pppoe_hdr)));
100 #define IS_PPPOE_IP(skb) \
101 (skb->protocol == htons(ETH_P_PPP_SES) && \
102 pppoe_proto(skb) == htons(PPP_IP) && \
103 brnf_filter_pppoe_tagged)
105 #define IS_PPPOE_IPV6(skb) \
106 (skb->protocol == htons(ETH_P_PPP_SES) && \
107 pppoe_proto(skb) == htons(PPP_IPV6) && \
108 brnf_filter_pppoe_tagged)
110 /* largest possible L2 header, see br_nf_dev_queue_xmit() */
111 #define NF_BRIDGE_MAX_MAC_HEADER_LENGTH (PPPOE_SES_HLEN + ETH_HLEN)
113 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4)
114 struct brnf_frag_data {
115 char mac[NF_BRIDGE_MAX_MAC_HEADER_LENGTH];
120 static DEFINE_PER_CPU(struct brnf_frag_data, brnf_frag_data_storage);
123 static struct nf_bridge_info *nf_bridge_info_get(const struct sk_buff *skb)
125 return skb->nf_bridge;
128 static void nf_bridge_info_free(struct sk_buff *skb)
130 if (skb->nf_bridge) {
131 nf_bridge_put(skb->nf_bridge);
132 skb->nf_bridge = NULL;
136 static inline struct rtable *bridge_parent_rtable(const struct net_device *dev)
138 struct net_bridge_port *port;
140 port = br_port_get_rcu(dev);
141 return port ? &port->br->fake_rtable : NULL;
144 static inline struct net_device *bridge_parent(const struct net_device *dev)
146 struct net_bridge_port *port;
148 port = br_port_get_rcu(dev);
149 return port ? port->br->dev : NULL;
152 static inline struct nf_bridge_info *nf_bridge_alloc(struct sk_buff *skb)
154 skb->nf_bridge = kzalloc(sizeof(struct nf_bridge_info), GFP_ATOMIC);
155 if (likely(skb->nf_bridge))
156 atomic_set(&(skb->nf_bridge->use), 1);
158 return skb->nf_bridge;
161 static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
163 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
165 if (atomic_read(&nf_bridge->use) > 1) {
166 struct nf_bridge_info *tmp = nf_bridge_alloc(skb);
169 memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info));
170 atomic_set(&tmp->use, 1);
172 nf_bridge_put(nf_bridge);
178 static unsigned int nf_bridge_encap_header_len(const struct sk_buff *skb)
180 switch (skb->protocol) {
181 case __cpu_to_be16(ETH_P_8021Q):
183 case __cpu_to_be16(ETH_P_PPP_SES):
184 return PPPOE_SES_HLEN;
190 static inline void nf_bridge_push_encap_header(struct sk_buff *skb)
192 unsigned int len = nf_bridge_encap_header_len(skb);
195 skb->network_header -= len;
198 static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
200 unsigned int len = nf_bridge_encap_header_len(skb);
203 skb->network_header += len;
206 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
208 unsigned int len = nf_bridge_encap_header_len(skb);
210 skb_pull_rcsum(skb, len);
211 skb->network_header += len;
214 /* When handing a packet over to the IP layer
215 * check whether we have a skb that is in the
219 static int br_validate_ipv4(struct sk_buff *skb)
221 const struct iphdr *iph;
222 struct net_device *dev = skb->dev;
225 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
230 /* Basic sanity checks */
231 if (iph->ihl < 5 || iph->version != 4)
234 if (!pskb_may_pull(skb, iph->ihl*4))
238 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
241 len = ntohs(iph->tot_len);
242 if (skb->len < len) {
243 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INTRUNCATEDPKTS);
245 } else if (len < (iph->ihl*4))
248 if (pskb_trim_rcsum(skb, len)) {
249 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
253 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
254 /* We should really parse IP options here but until
255 * somebody who actually uses IP options complains to
256 * us we'll just silently ignore the options because
262 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
267 /* We only check the length. A bridge shouldn't do any hop-by-hop stuff
270 static int check_hbh_len(struct sk_buff *skb)
272 unsigned char *raw = (u8 *)(ipv6_hdr(skb) + 1);
274 const unsigned char *nh = skb_network_header(skb);
276 int len = (raw[1] + 1) << 3;
278 if ((raw + len) - skb->data > skb_headlen(skb))
285 int optlen = nh[off + 1] + 2;
296 if (nh[off + 1] != 4 || (off & 3) != 2)
298 pkt_len = ntohl(*(__be32 *)(nh + off + 2));
299 if (pkt_len <= IPV6_MAXPLEN ||
300 ipv6_hdr(skb)->payload_len)
302 if (pkt_len > skb->len - sizeof(struct ipv6hdr))
304 if (pskb_trim_rcsum(skb,
305 pkt_len + sizeof(struct ipv6hdr)))
307 nh = skb_network_header(skb);
323 static void nf_bridge_update_protocol(struct sk_buff *skb)
325 switch (skb->nf_bridge->orig_proto) {
326 case BRNF_PROTO_8021Q:
327 skb->protocol = htons(ETH_P_8021Q);
329 case BRNF_PROTO_PPPOE:
330 skb->protocol = htons(ETH_P_PPP_SES);
332 case BRNF_PROTO_UNCHANGED:
337 /* Obtain the correct destination MAC address, while preserving the original
338 * source MAC address. If we already know this address, we just copy it. If we
339 * don't, we use the neighbour framework to find out. In both cases, we make
340 * sure that br_handle_frame_finish() is called afterwards.
342 static int br_nf_pre_routing_finish_bridge(struct sock *sk, struct sk_buff *skb)
344 struct neighbour *neigh;
345 struct dst_entry *dst;
347 skb->dev = bridge_parent(skb->dev);
351 neigh = dst_neigh_lookup_skb(dst, skb);
353 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
356 if (neigh->hh.hh_len) {
357 neigh_hh_bridge(&neigh->hh, skb);
358 skb->dev = nf_bridge->physindev;
359 ret = br_handle_frame_finish(sk, skb);
361 /* the neighbour function below overwrites the complete
362 * MAC header, so we save the Ethernet source address and
365 skb_copy_from_linear_data_offset(skb,
366 -(ETH_HLEN-ETH_ALEN),
367 nf_bridge->neigh_header,
369 /* tell br_dev_xmit to continue with forwarding */
370 nf_bridge->mask |= BRNF_BRIDGED_DNAT;
371 /* FIXME Need to refragment */
372 ret = neigh->output(neigh, skb);
374 neigh_release(neigh);
382 static bool daddr_was_changed(const struct sk_buff *skb,
383 const struct nf_bridge_info *nf_bridge)
385 switch (skb->protocol) {
386 case htons(ETH_P_IP):
387 return ip_hdr(skb)->daddr != nf_bridge->ipv4_daddr;
388 case htons(ETH_P_IPV6):
389 return memcmp(&nf_bridge->ipv6_daddr, &ipv6_hdr(skb)->daddr,
390 sizeof(ipv6_hdr(skb)->daddr)) != 0;
396 /* PF_BRIDGE/PRE_ROUTING: Undo the changes made for ip6tables
397 * PREROUTING and continue the bridge PRE_ROUTING hook. See comment
398 * for br_nf_pre_routing_finish(), same logic is used here but
399 * equivalent IPv6 function ip6_route_input() called indirectly.
401 static int br_nf_pre_routing_finish_ipv6(struct sock *sk, struct sk_buff *skb)
403 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
405 struct net_device *dev = skb->dev;
406 const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops();
408 if (nf_bridge->pkt_otherhost) {
409 skb->pkt_type = PACKET_OTHERHOST;
410 nf_bridge->pkt_otherhost = false;
412 nf_bridge->mask &= ~BRNF_NF_BRIDGE_PREROUTING;
413 if (daddr_was_changed(skb, nf_bridge)) {
415 v6ops->route_input(skb);
417 if (skb_dst(skb)->error) {
422 if (skb_dst(skb)->dev == dev) {
423 skb->dev = nf_bridge->physindev;
424 nf_bridge_update_protocol(skb);
425 nf_bridge_push_encap_header(skb);
426 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING,
427 sk, skb, skb->dev, NULL,
428 br_nf_pre_routing_finish_bridge,
432 ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr);
433 skb->pkt_type = PACKET_HOST;
435 rt = bridge_parent_rtable(nf_bridge->physindev);
440 skb_dst_set_noref(skb, &rt->dst);
443 skb->dev = nf_bridge->physindev;
444 nf_bridge_update_protocol(skb);
445 nf_bridge_push_encap_header(skb);
446 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, sk, skb,
448 br_handle_frame_finish, 1);
453 /* This requires some explaining. If DNAT has taken place,
454 * we will need to fix up the destination Ethernet address.
455 * This is also true when SNAT takes place (for the reply direction).
457 * There are two cases to consider:
458 * 1. The packet was DNAT'ed to a device in the same bridge
459 * port group as it was received on. We can still bridge
461 * 2. The packet was DNAT'ed to a different device, either
462 * a non-bridged device or another bridge port group.
463 * The packet will need to be routed.
465 * The correct way of distinguishing between these two cases is to
466 * call ip_route_input() and to look at skb->dst->dev, which is
467 * changed to the destination device if ip_route_input() succeeds.
469 * Let's first consider the case that ip_route_input() succeeds:
471 * If the output device equals the logical bridge device the packet
472 * came in on, we can consider this bridging. The corresponding MAC
473 * address will be obtained in br_nf_pre_routing_finish_bridge.
474 * Otherwise, the packet is considered to be routed and we just
475 * change the destination MAC address so that the packet will
476 * later be passed up to the IP stack to be routed. For a redirected
477 * packet, ip_route_input() will give back the localhost as output device,
478 * which differs from the bridge device.
480 * Let's now consider the case that ip_route_input() fails:
482 * This can be because the destination address is martian, in which case
483 * the packet will be dropped.
484 * If IP forwarding is disabled, ip_route_input() will fail, while
485 * ip_route_output_key() can return success. The source
486 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
487 * thinks we're handling a locally generated packet and won't care
488 * if IP forwarding is enabled. If the output device equals the logical bridge
489 * device, we proceed as if ip_route_input() succeeded. If it differs from the
490 * logical bridge port or if ip_route_output_key() fails we drop the packet.
492 static int br_nf_pre_routing_finish(struct sock *sk, struct sk_buff *skb)
494 struct net_device *dev = skb->dev;
495 struct iphdr *iph = ip_hdr(skb);
496 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
500 nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
502 if (nf_bridge->pkt_otherhost) {
503 skb->pkt_type = PACKET_OTHERHOST;
504 nf_bridge->pkt_otherhost = false;
506 nf_bridge->mask &= ~BRNF_NF_BRIDGE_PREROUTING;
507 if (daddr_was_changed(skb, nf_bridge)) {
508 if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
509 struct in_device *in_dev = __in_dev_get_rcu(dev);
511 /* If err equals -EHOSTUNREACH the error is due to a
512 * martian destination or due to the fact that
513 * forwarding is disabled. For most martian packets,
514 * ip_route_output_key() will fail. It won't fail for 2 types of
515 * martian destinations: loopback destinations and destination
516 * 0.0.0.0. In both cases the packet will be dropped because the
517 * destination is the loopback device and not the bridge. */
518 if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
521 rt = ip_route_output(dev_net(dev), iph->daddr, 0,
522 RT_TOS(iph->tos), 0);
524 /* - Bridged-and-DNAT'ed traffic doesn't
525 * require ip_forwarding. */
526 if (rt->dst.dev == dev) {
527 skb_dst_set(skb, &rt->dst);
536 if (skb_dst(skb)->dev == dev) {
538 skb->dev = nf_bridge->physindev;
539 nf_bridge_update_protocol(skb);
540 nf_bridge_push_encap_header(skb);
541 NF_HOOK_THRESH(NFPROTO_BRIDGE,
543 sk, skb, skb->dev, NULL,
544 br_nf_pre_routing_finish_bridge,
548 ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr);
549 skb->pkt_type = PACKET_HOST;
552 rt = bridge_parent_rtable(nf_bridge->physindev);
557 skb_dst_set_noref(skb, &rt->dst);
560 skb->dev = nf_bridge->physindev;
561 nf_bridge_update_protocol(skb);
562 nf_bridge_push_encap_header(skb);
563 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, sk, skb,
565 br_handle_frame_finish, 1);
570 static struct net_device *brnf_get_logical_dev(struct sk_buff *skb, const struct net_device *dev)
572 struct net_device *vlan, *br;
574 br = bridge_parent(dev);
575 if (brnf_pass_vlan_indev == 0 || !skb_vlan_tag_present(skb))
578 vlan = __vlan_find_dev_deep_rcu(br, skb->vlan_proto,
579 skb_vlan_tag_get(skb) & VLAN_VID_MASK);
581 return vlan ? vlan : br;
584 /* Some common code for IPv4/IPv6 */
585 static struct net_device *setup_pre_routing(struct sk_buff *skb)
587 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
589 if (skb->pkt_type == PACKET_OTHERHOST) {
590 skb->pkt_type = PACKET_HOST;
591 nf_bridge->pkt_otherhost = true;
594 nf_bridge->mask |= BRNF_NF_BRIDGE_PREROUTING;
595 nf_bridge->physindev = skb->dev;
596 skb->dev = brnf_get_logical_dev(skb, skb->dev);
598 if (skb->protocol == htons(ETH_P_8021Q))
599 nf_bridge->orig_proto = BRNF_PROTO_8021Q;
600 else if (skb->protocol == htons(ETH_P_PPP_SES))
601 nf_bridge->orig_proto = BRNF_PROTO_PPPOE;
603 /* Must drop socket now because of tproxy. */
608 /* Replicate the checks that IPv6 does on packet reception and pass the packet
609 * to ip6tables, which doesn't support NAT, so things are fairly simple. */
610 static unsigned int br_nf_pre_routing_ipv6(const struct nf_hook_ops *ops,
612 const struct nf_hook_state *state)
614 struct nf_bridge_info *nf_bridge;
615 const struct ipv6hdr *hdr;
618 if (skb->len < sizeof(struct ipv6hdr))
621 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
626 if (hdr->version != 6)
629 pkt_len = ntohs(hdr->payload_len);
631 if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
632 if (pkt_len + sizeof(struct ipv6hdr) > skb->len)
634 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
637 if (hdr->nexthdr == NEXTHDR_HOP && check_hbh_len(skb))
640 nf_bridge_put(skb->nf_bridge);
641 if (!nf_bridge_alloc(skb))
643 if (!setup_pre_routing(skb))
646 nf_bridge = nf_bridge_info_get(skb);
647 nf_bridge->ipv6_daddr = ipv6_hdr(skb)->daddr;
649 skb->protocol = htons(ETH_P_IPV6);
650 NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, state->sk, skb,
652 br_nf_pre_routing_finish_ipv6);
657 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
658 * Replicate the checks that IPv4 does on packet reception.
659 * Set skb->dev to the bridge device (i.e. parent of the
660 * receiving device) to make netfilter happy, the REDIRECT
661 * target in particular. Save the original destination IP
662 * address to be able to detect DNAT afterwards. */
663 static unsigned int br_nf_pre_routing(const struct nf_hook_ops *ops,
665 const struct nf_hook_state *state)
667 struct nf_bridge_info *nf_bridge;
668 struct net_bridge_port *p;
669 struct net_bridge *br;
670 __u32 len = nf_bridge_encap_header_len(skb);
672 if (unlikely(!pskb_may_pull(skb, len)))
675 p = br_port_get_rcu(state->in);
680 if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb)) {
681 if (!brnf_call_ip6tables && !br->nf_call_ip6tables)
684 nf_bridge_pull_encap_header_rcsum(skb);
685 return br_nf_pre_routing_ipv6(ops, skb, state);
688 if (!brnf_call_iptables && !br->nf_call_iptables)
691 if (!IS_IP(skb) && !IS_VLAN_IP(skb) && !IS_PPPOE_IP(skb))
694 nf_bridge_pull_encap_header_rcsum(skb);
696 if (br_validate_ipv4(skb))
699 nf_bridge_put(skb->nf_bridge);
700 if (!nf_bridge_alloc(skb))
702 if (!setup_pre_routing(skb))
705 nf_bridge = nf_bridge_info_get(skb);
706 nf_bridge->ipv4_daddr = ip_hdr(skb)->daddr;
708 skb->protocol = htons(ETH_P_IP);
710 NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, state->sk, skb,
712 br_nf_pre_routing_finish);
718 /* PF_BRIDGE/LOCAL_IN ************************************************/
719 /* The packet is locally destined, which requires a real
720 * dst_entry, so detach the fake one. On the way up, the
721 * packet would pass through PRE_ROUTING again (which already
722 * took place when the packet entered the bridge), but we
723 * register an IPv4 PRE_ROUTING 'sabotage' hook that will
724 * prevent this from happening. */
725 static unsigned int br_nf_local_in(const struct nf_hook_ops *ops,
727 const struct nf_hook_state *state)
729 br_drop_fake_rtable(skb);
733 /* PF_BRIDGE/FORWARD *************************************************/
734 static int br_nf_forward_finish(struct sock *sk, struct sk_buff *skb)
736 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
737 struct net_device *in;
739 if (!IS_ARP(skb) && !IS_VLAN_ARP(skb)) {
741 if (skb->protocol == htons(ETH_P_IP)) {
742 nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
745 in = nf_bridge->physindev;
746 if (nf_bridge->pkt_otherhost) {
747 skb->pkt_type = PACKET_OTHERHOST;
748 nf_bridge->pkt_otherhost = false;
750 nf_bridge_update_protocol(skb);
752 in = *((struct net_device **)(skb->cb));
754 nf_bridge_push_encap_header(skb);
756 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_FORWARD, sk, skb,
757 in, skb->dev, br_forward_finish, 1);
762 /* This is the 'purely bridged' case. For IP, we pass the packet to
763 * netfilter with indev and outdev set to the bridge device,
764 * but we are still able to filter on the 'real' indev/outdev
765 * because of the physdev module. For ARP, indev and outdev are the
767 static unsigned int br_nf_forward_ip(const struct nf_hook_ops *ops,
769 const struct nf_hook_state *state)
771 struct nf_bridge_info *nf_bridge;
772 struct net_device *parent;
778 /* Need exclusive nf_bridge_info since we might have multiple
779 * different physoutdevs. */
780 if (!nf_bridge_unshare(skb))
783 nf_bridge = nf_bridge_info_get(skb);
787 parent = bridge_parent(state->out);
791 if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
793 else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
798 nf_bridge_pull_encap_header(skb);
800 if (skb->pkt_type == PACKET_OTHERHOST) {
801 skb->pkt_type = PACKET_HOST;
802 nf_bridge->pkt_otherhost = true;
805 if (pf == NFPROTO_IPV4) {
806 if (br_validate_ipv4(skb))
808 IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
811 nf_bridge->physoutdev = skb->dev;
812 if (pf == NFPROTO_IPV4)
813 skb->protocol = htons(ETH_P_IP);
815 skb->protocol = htons(ETH_P_IPV6);
817 NF_HOOK(pf, NF_INET_FORWARD, NULL, skb,
818 brnf_get_logical_dev(skb, state->in),
819 parent, br_nf_forward_finish);
824 static unsigned int br_nf_forward_arp(const struct nf_hook_ops *ops,
826 const struct nf_hook_state *state)
828 struct net_bridge_port *p;
829 struct net_bridge *br;
830 struct net_device **d = (struct net_device **)(skb->cb);
832 p = br_port_get_rcu(state->out);
837 if (!brnf_call_arptables && !br->nf_call_arptables)
841 if (!IS_VLAN_ARP(skb))
843 nf_bridge_pull_encap_header(skb);
846 if (arp_hdr(skb)->ar_pln != 4) {
847 if (IS_VLAN_ARP(skb))
848 nf_bridge_push_encap_header(skb);
852 NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, state->sk, skb,
853 state->in, state->out, br_nf_forward_finish);
858 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4)
859 static int br_nf_push_frag_xmit(struct sock *sk, struct sk_buff *skb)
861 struct brnf_frag_data *data;
864 data = this_cpu_ptr(&brnf_frag_data_storage);
865 err = skb_cow_head(skb, data->size);
872 skb_copy_to_linear_data_offset(skb, -data->size, data->mac, data->size);
873 __skb_push(skb, data->encap_size);
875 nf_bridge_info_free(skb);
876 return br_dev_queue_push_xmit(sk, skb);
879 static int br_nf_ip_fragment(struct sock *sk, struct sk_buff *skb,
880 int (*output)(struct sock *, struct sk_buff *))
882 unsigned int mtu = ip_skb_dst_mtu(skb);
883 struct iphdr *iph = ip_hdr(skb);
884 struct rtable *rt = skb_rtable(skb);
885 struct net_device *dev = rt->dst.dev;
887 if (unlikely(((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) ||
888 (IPCB(skb)->frag_max_size &&
889 IPCB(skb)->frag_max_size > mtu))) {
890 IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
895 return ip_do_fragment(sk, skb, output);
898 static int br_nf_dev_queue_xmit(struct sock *sk, struct sk_buff *skb)
901 struct nf_bridge_info *nf_bridge;
902 unsigned int mtu_reserved;
904 if (skb_is_gso(skb) || skb->protocol != htons(ETH_P_IP)) {
905 nf_bridge_info_free(skb);
906 return br_dev_queue_push_xmit(sk, skb);
909 mtu_reserved = nf_bridge_mtu_reduction(skb);
910 nf_bridge = nf_bridge_info_get(skb);
911 /* This is wrong! We should preserve the original fragment
912 * boundaries by preserving frag_list rather than refragmenting.
914 if (skb->len + mtu_reserved > skb->dev->mtu) {
915 struct brnf_frag_data *data;
917 if (br_validate_ipv4(skb))
920 IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
922 nf_bridge_update_protocol(skb);
924 data = this_cpu_ptr(&brnf_frag_data_storage);
925 data->encap_size = nf_bridge_encap_header_len(skb);
926 data->size = ETH_HLEN + data->encap_size;
928 skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
931 ret = br_nf_ip_fragment(sk, skb, br_nf_push_frag_xmit);
933 nf_bridge_info_free(skb);
934 ret = br_dev_queue_push_xmit(sk, skb);
940 static int br_nf_dev_queue_xmit(struct sock *sk, struct sk_buff *skb)
942 nf_bridge_info_free(skb);
943 return br_dev_queue_push_xmit(sk, skb);
947 /* PF_BRIDGE/POST_ROUTING ********************************************/
948 static unsigned int br_nf_post_routing(const struct nf_hook_ops *ops,
950 const struct nf_hook_state *state)
952 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
953 struct net_device *realoutdev = bridge_parent(skb->dev);
956 /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
957 * on a bridge, but was delivered locally and is now being routed:
959 * POST_ROUTING was already invoked from the ip stack.
961 if (!nf_bridge || !nf_bridge->physoutdev)
967 if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
969 else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
974 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
975 * about the value of skb->pkt_type. */
976 if (skb->pkt_type == PACKET_OTHERHOST) {
977 skb->pkt_type = PACKET_HOST;
978 nf_bridge->pkt_otherhost = true;
981 nf_bridge_pull_encap_header(skb);
982 if (pf == NFPROTO_IPV4)
983 skb->protocol = htons(ETH_P_IP);
985 skb->protocol = htons(ETH_P_IPV6);
987 NF_HOOK(pf, NF_INET_POST_ROUTING, state->sk, skb,
989 br_nf_dev_queue_xmit);
994 /* IP/SABOTAGE *****************************************************/
995 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
996 * for the second time. */
997 static unsigned int ip_sabotage_in(const struct nf_hook_ops *ops,
999 const struct nf_hook_state *state)
1001 if (skb->nf_bridge &&
1002 !(skb->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)) {
1009 /* This is called when br_netfilter has called into iptables/netfilter,
1010 * and DNAT has taken place on a bridge-forwarded packet.
1012 * neigh->output has created a new MAC header, with local br0 MAC
1015 * This restores the original MAC saddr of the bridged packet
1016 * before invoking bridge forward logic to transmit the packet.
1018 static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff *skb)
1020 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
1022 skb_pull(skb, ETH_HLEN);
1023 nf_bridge->mask &= ~BRNF_BRIDGED_DNAT;
1025 BUILD_BUG_ON(sizeof(nf_bridge->neigh_header) != (ETH_HLEN - ETH_ALEN));
1027 skb_copy_to_linear_data_offset(skb, -(ETH_HLEN - ETH_ALEN),
1028 nf_bridge->neigh_header,
1029 ETH_HLEN - ETH_ALEN);
1030 skb->dev = nf_bridge->physindev;
1032 nf_bridge->physoutdev = NULL;
1033 br_handle_frame_finish(NULL, skb);
1036 static int br_nf_dev_xmit(struct sk_buff *skb)
1038 if (skb->nf_bridge && (skb->nf_bridge->mask & BRNF_BRIDGED_DNAT)) {
1039 br_nf_pre_routing_finish_bridge_slow(skb);
1045 static const struct nf_br_ops br_ops = {
1046 .br_dev_xmit_hook = br_nf_dev_xmit,
1049 void br_netfilter_enable(void)
1052 EXPORT_SYMBOL_GPL(br_netfilter_enable);
1054 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
1055 * br_dev_queue_push_xmit is called afterwards */
1056 static struct nf_hook_ops br_nf_ops[] __read_mostly = {
1058 .hook = br_nf_pre_routing,
1059 .owner = THIS_MODULE,
1060 .pf = NFPROTO_BRIDGE,
1061 .hooknum = NF_BR_PRE_ROUTING,
1062 .priority = NF_BR_PRI_BRNF,
1065 .hook = br_nf_local_in,
1066 .owner = THIS_MODULE,
1067 .pf = NFPROTO_BRIDGE,
1068 .hooknum = NF_BR_LOCAL_IN,
1069 .priority = NF_BR_PRI_BRNF,
1072 .hook = br_nf_forward_ip,
1073 .owner = THIS_MODULE,
1074 .pf = NFPROTO_BRIDGE,
1075 .hooknum = NF_BR_FORWARD,
1076 .priority = NF_BR_PRI_BRNF - 1,
1079 .hook = br_nf_forward_arp,
1080 .owner = THIS_MODULE,
1081 .pf = NFPROTO_BRIDGE,
1082 .hooknum = NF_BR_FORWARD,
1083 .priority = NF_BR_PRI_BRNF,
1086 .hook = br_nf_post_routing,
1087 .owner = THIS_MODULE,
1088 .pf = NFPROTO_BRIDGE,
1089 .hooknum = NF_BR_POST_ROUTING,
1090 .priority = NF_BR_PRI_LAST,
1093 .hook = ip_sabotage_in,
1094 .owner = THIS_MODULE,
1096 .hooknum = NF_INET_PRE_ROUTING,
1097 .priority = NF_IP_PRI_FIRST,
1100 .hook = ip_sabotage_in,
1101 .owner = THIS_MODULE,
1103 .hooknum = NF_INET_PRE_ROUTING,
1104 .priority = NF_IP6_PRI_FIRST,
1108 #ifdef CONFIG_SYSCTL
1110 int brnf_sysctl_call_tables(struct ctl_table *ctl, int write,
1111 void __user *buffer, size_t *lenp, loff_t *ppos)
1115 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1117 if (write && *(int *)(ctl->data))
1118 *(int *)(ctl->data) = 1;
1122 static struct ctl_table brnf_table[] = {
1124 .procname = "bridge-nf-call-arptables",
1125 .data = &brnf_call_arptables,
1126 .maxlen = sizeof(int),
1128 .proc_handler = brnf_sysctl_call_tables,
1131 .procname = "bridge-nf-call-iptables",
1132 .data = &brnf_call_iptables,
1133 .maxlen = sizeof(int),
1135 .proc_handler = brnf_sysctl_call_tables,
1138 .procname = "bridge-nf-call-ip6tables",
1139 .data = &brnf_call_ip6tables,
1140 .maxlen = sizeof(int),
1142 .proc_handler = brnf_sysctl_call_tables,
1145 .procname = "bridge-nf-filter-vlan-tagged",
1146 .data = &brnf_filter_vlan_tagged,
1147 .maxlen = sizeof(int),
1149 .proc_handler = brnf_sysctl_call_tables,
1152 .procname = "bridge-nf-filter-pppoe-tagged",
1153 .data = &brnf_filter_pppoe_tagged,
1154 .maxlen = sizeof(int),
1156 .proc_handler = brnf_sysctl_call_tables,
1159 .procname = "bridge-nf-pass-vlan-input-dev",
1160 .data = &brnf_pass_vlan_indev,
1161 .maxlen = sizeof(int),
1163 .proc_handler = brnf_sysctl_call_tables,
1169 static int __init br_netfilter_init(void)
1173 ret = nf_register_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
1177 #ifdef CONFIG_SYSCTL
1178 brnf_sysctl_header = register_net_sysctl(&init_net, "net/bridge", brnf_table);
1179 if (brnf_sysctl_header == NULL) {
1181 "br_netfilter: can't register to sysctl.\n");
1182 nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
1186 RCU_INIT_POINTER(nf_br_ops, &br_ops);
1187 printk(KERN_NOTICE "Bridge firewalling registered\n");
1191 static void __exit br_netfilter_fini(void)
1193 RCU_INIT_POINTER(nf_br_ops, NULL);
1194 nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
1195 #ifdef CONFIG_SYSCTL
1196 unregister_net_sysctl_table(brnf_sysctl_header);
1200 module_init(br_netfilter_init);
1201 module_exit(br_netfilter_fini);
1203 MODULE_LICENSE("GPL");
1204 MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
1205 MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
1206 MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");