netfilter: bridge: rename br_netfilter.c to br_netfilter_hooks.c
authorPablo Neira Ayuso <pablo@netfilter.org>
Tue, 16 Jun 2015 11:38:26 +0000 (13:38 +0200)
committerPablo Neira Ayuso <pablo@netfilter.org>
Thu, 18 Jun 2015 11:05:53 +0000 (13:05 +0200)
To prepare separation of the IPv6 code into different file.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
net/bridge/Makefile
net/bridge/br_netfilter.c [deleted file]
net/bridge/br_netfilter_hooks.c [new file with mode: 0644]

index fd7ee03c59b3f6c891fe9b5020b08af389d3e814..c52577ac718e8fd746f2821d6aba681b8b4eefdc 100644 (file)
@@ -12,6 +12,7 @@ bridge-$(CONFIG_SYSFS) += br_sysfs_if.o br_sysfs_br.o
 
 bridge-$(subst m,y,$(CONFIG_BRIDGE_NETFILTER)) += br_nf_core.o
 
+br_netfilter-y := br_netfilter_hooks.o
 obj-$(CONFIG_BRIDGE_NETFILTER) += br_netfilter.o
 
 bridge-$(CONFIG_BRIDGE_IGMP_SNOOPING) += br_multicast.o br_mdb.o
diff --git a/net/bridge/br_netfilter.c b/net/bridge/br_netfilter.c
deleted file mode 100644 (file)
index e4e5f2f..0000000
+++ /dev/null
@@ -1,1282 +0,0 @@
-/*
- *     Handle firewalling
- *     Linux ethernet bridge
- *
- *     Authors:
- *     Lennert Buytenhek               <buytenh@gnu.org>
- *     Bart De Schuymer                <bdschuym@pandora.be>
- *
- *     This program is free software; you can redistribute it and/or
- *     modify it under the terms of the GNU General Public License
- *     as published by the Free Software Foundation; either version
- *     2 of the License, or (at your option) any later version.
- *
- *     Lennert dedicates this file to Kerstin Wurdinger.
- */
-
-#include <linux/module.h>
-#include <linux/kernel.h>
-#include <linux/slab.h>
-#include <linux/ip.h>
-#include <linux/netdevice.h>
-#include <linux/skbuff.h>
-#include <linux/if_arp.h>
-#include <linux/if_ether.h>
-#include <linux/if_vlan.h>
-#include <linux/if_pppox.h>
-#include <linux/ppp_defs.h>
-#include <linux/netfilter_bridge.h>
-#include <linux/netfilter_ipv4.h>
-#include <linux/netfilter_ipv6.h>
-#include <linux/netfilter_arp.h>
-#include <linux/in_route.h>
-#include <linux/inetdevice.h>
-
-#include <net/ip.h>
-#include <net/ipv6.h>
-#include <net/addrconf.h>
-#include <net/route.h>
-#include <net/netfilter/br_netfilter.h>
-
-#include <asm/uaccess.h>
-#include "br_private.h"
-#ifdef CONFIG_SYSCTL
-#include <linux/sysctl.h>
-#endif
-
-#ifdef CONFIG_SYSCTL
-static struct ctl_table_header *brnf_sysctl_header;
-static int brnf_call_iptables __read_mostly = 1;
-static int brnf_call_ip6tables __read_mostly = 1;
-static int brnf_call_arptables __read_mostly = 1;
-static int brnf_filter_vlan_tagged __read_mostly = 0;
-static int brnf_filter_pppoe_tagged __read_mostly = 0;
-static int brnf_pass_vlan_indev __read_mostly = 0;
-#else
-#define brnf_call_iptables 1
-#define brnf_call_ip6tables 1
-#define brnf_call_arptables 1
-#define brnf_filter_vlan_tagged 0
-#define brnf_filter_pppoe_tagged 0
-#define brnf_pass_vlan_indev 0
-#endif
-
-#define IS_IP(skb) \
-       (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP))
-
-#define IS_IPV6(skb) \
-       (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6))
-
-#define IS_ARP(skb) \
-       (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP))
-
-static inline __be16 vlan_proto(const struct sk_buff *skb)
-{
-       if (skb_vlan_tag_present(skb))
-               return skb->protocol;
-       else if (skb->protocol == htons(ETH_P_8021Q))
-               return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
-       else
-               return 0;
-}
-
-#define IS_VLAN_IP(skb) \
-       (vlan_proto(skb) == htons(ETH_P_IP) && \
-        brnf_filter_vlan_tagged)
-
-#define IS_VLAN_IPV6(skb) \
-       (vlan_proto(skb) == htons(ETH_P_IPV6) && \
-        brnf_filter_vlan_tagged)
-
-#define IS_VLAN_ARP(skb) \
-       (vlan_proto(skb) == htons(ETH_P_ARP) && \
-        brnf_filter_vlan_tagged)
-
-static inline __be16 pppoe_proto(const struct sk_buff *skb)
-{
-       return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
-                           sizeof(struct pppoe_hdr)));
-}
-
-#define IS_PPPOE_IP(skb) \
-       (skb->protocol == htons(ETH_P_PPP_SES) && \
-        pppoe_proto(skb) == htons(PPP_IP) && \
-        brnf_filter_pppoe_tagged)
-
-#define IS_PPPOE_IPV6(skb) \
-       (skb->protocol == htons(ETH_P_PPP_SES) && \
-        pppoe_proto(skb) == htons(PPP_IPV6) && \
-        brnf_filter_pppoe_tagged)
-
-/* largest possible L2 header, see br_nf_dev_queue_xmit() */
-#define NF_BRIDGE_MAX_MAC_HEADER_LENGTH (PPPOE_SES_HLEN + ETH_HLEN)
-
-#if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4)
-struct brnf_frag_data {
-       char mac[NF_BRIDGE_MAX_MAC_HEADER_LENGTH];
-       u8 encap_size;
-       u8 size;
-       u16 vlan_tci;
-       __be16 vlan_proto;
-};
-
-static DEFINE_PER_CPU(struct brnf_frag_data, brnf_frag_data_storage);
-#endif
-
-static struct nf_bridge_info *nf_bridge_info_get(const struct sk_buff *skb)
-{
-       return skb->nf_bridge;
-}
-
-static void nf_bridge_info_free(struct sk_buff *skb)
-{
-       if (skb->nf_bridge) {
-               nf_bridge_put(skb->nf_bridge);
-               skb->nf_bridge = NULL;
-       }
-}
-
-static inline struct rtable *bridge_parent_rtable(const struct net_device *dev)
-{
-       struct net_bridge_port *port;
-
-       port = br_port_get_rcu(dev);
-       return port ? &port->br->fake_rtable : NULL;
-}
-
-static inline struct net_device *bridge_parent(const struct net_device *dev)
-{
-       struct net_bridge_port *port;
-
-       port = br_port_get_rcu(dev);
-       return port ? port->br->dev : NULL;
-}
-
-static inline struct nf_bridge_info *nf_bridge_alloc(struct sk_buff *skb)
-{
-       skb->nf_bridge = kzalloc(sizeof(struct nf_bridge_info), GFP_ATOMIC);
-       if (likely(skb->nf_bridge))
-               atomic_set(&(skb->nf_bridge->use), 1);
-
-       return skb->nf_bridge;
-}
-
-static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
-{
-       struct nf_bridge_info *nf_bridge = skb->nf_bridge;
-
-       if (atomic_read(&nf_bridge->use) > 1) {
-               struct nf_bridge_info *tmp = nf_bridge_alloc(skb);
-
-               if (tmp) {
-                       memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info));
-                       atomic_set(&tmp->use, 1);
-               }
-               nf_bridge_put(nf_bridge);
-               nf_bridge = tmp;
-       }
-       return nf_bridge;
-}
-
-static unsigned int nf_bridge_encap_header_len(const struct sk_buff *skb)
-{
-       switch (skb->protocol) {
-       case __cpu_to_be16(ETH_P_8021Q):
-               return VLAN_HLEN;
-       case __cpu_to_be16(ETH_P_PPP_SES):
-               return PPPOE_SES_HLEN;
-       default:
-               return 0;
-       }
-}
-
-static inline void nf_bridge_push_encap_header(struct sk_buff *skb)
-{
-       unsigned int len = nf_bridge_encap_header_len(skb);
-
-       skb_push(skb, len);
-       skb->network_header -= len;
-}
-
-static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
-{
-       unsigned int len = nf_bridge_encap_header_len(skb);
-
-       skb_pull(skb, len);
-       skb->network_header += len;
-}
-
-static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
-{
-       unsigned int len = nf_bridge_encap_header_len(skb);
-
-       skb_pull_rcsum(skb, len);
-       skb->network_header += len;
-}
-
-/* When handing a packet over to the IP layer
- * check whether we have a skb that is in the
- * expected format
- */
-
-static int br_validate_ipv4(struct sk_buff *skb)
-{
-       const struct iphdr *iph;
-       struct net_device *dev = skb->dev;
-       u32 len;
-
-       if (!pskb_may_pull(skb, sizeof(struct iphdr)))
-               goto inhdr_error;
-
-       iph = ip_hdr(skb);
-
-       /* Basic sanity checks */
-       if (iph->ihl < 5 || iph->version != 4)
-               goto inhdr_error;
-
-       if (!pskb_may_pull(skb, iph->ihl*4))
-               goto inhdr_error;
-
-       iph = ip_hdr(skb);
-       if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
-               goto inhdr_error;
-
-       len = ntohs(iph->tot_len);
-       if (skb->len < len) {
-               IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INTRUNCATEDPKTS);
-               goto drop;
-       } else if (len < (iph->ihl*4))
-               goto inhdr_error;
-
-       if (pskb_trim_rcsum(skb, len)) {
-               IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
-               goto drop;
-       }
-
-       memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
-       /* We should really parse IP options here but until
-        * somebody who actually uses IP options complains to
-        * us we'll just silently ignore the options because
-        * we're lazy!
-        */
-       return 0;
-
-inhdr_error:
-       IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
-drop:
-       return -1;
-}
-
-/* We only check the length. A bridge shouldn't do any hop-by-hop stuff
- * anyway
- */
-static int check_hbh_len(struct sk_buff *skb)
-{
-       unsigned char *raw = (u8 *)(ipv6_hdr(skb) + 1);
-       u32 pkt_len;
-       const unsigned char *nh = skb_network_header(skb);
-       int off = raw - nh;
-       int len = (raw[1] + 1) << 3;
-
-       if ((raw + len) - skb->data > skb_headlen(skb))
-               goto bad;
-
-       off += 2;
-       len -= 2;
-
-       while (len > 0) {
-               int optlen = nh[off + 1] + 2;
-
-               switch (nh[off]) {
-               case IPV6_TLV_PAD1:
-                       optlen = 1;
-                       break;
-
-               case IPV6_TLV_PADN:
-                       break;
-
-               case IPV6_TLV_JUMBO:
-                       if (nh[off + 1] != 4 || (off & 3) != 2)
-                               goto bad;
-                       pkt_len = ntohl(*(__be32 *)(nh + off + 2));
-                       if (pkt_len <= IPV6_MAXPLEN ||
-                           ipv6_hdr(skb)->payload_len)
-                               goto bad;
-                       if (pkt_len > skb->len - sizeof(struct ipv6hdr))
-                               goto bad;
-                       if (pskb_trim_rcsum(skb,
-                                           pkt_len + sizeof(struct ipv6hdr)))
-                               goto bad;
-                       nh = skb_network_header(skb);
-                       break;
-               default:
-                       if (optlen > len)
-                               goto bad;
-                       break;
-               }
-               off += optlen;
-               len -= optlen;
-       }
-       if (len == 0)
-               return 0;
-bad:
-       return -1;
-}
-
-/* Equivalent to br_validate_ipv4 for IPv6 */
-static int br_validate_ipv6(struct sk_buff *skb)
-{
-       const struct ipv6hdr *hdr;
-       struct net_device *dev = skb->dev;
-       struct inet6_dev *idev = in6_dev_get(skb->dev);
-       u32 pkt_len;
-       u8 ip6h_len = sizeof(struct ipv6hdr);
-
-       if (!pskb_may_pull(skb, ip6h_len))
-               goto inhdr_error;
-
-       if (skb->len < ip6h_len)
-               goto drop;
-
-       hdr = ipv6_hdr(skb);
-
-       if (hdr->version != 6)
-               goto inhdr_error;
-
-       pkt_len = ntohs(hdr->payload_len);
-
-       if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
-               if (pkt_len + ip6h_len > skb->len) {
-                       IP6_INC_STATS_BH(dev_net(dev), idev,
-                                        IPSTATS_MIB_INTRUNCATEDPKTS);
-                       goto drop;
-               }
-               if (pskb_trim_rcsum(skb, pkt_len + ip6h_len)) {
-                       IP6_INC_STATS_BH(dev_net(dev), idev,
-                                        IPSTATS_MIB_INDISCARDS);
-                       goto drop;
-               }
-       }
-       if (hdr->nexthdr == NEXTHDR_HOP && check_hbh_len(skb))
-               goto drop;
-
-       memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
-       /* No IP options in IPv6 header; however it should be
-        * checked if some next headers need special treatment
-        */
-       return 0;
-
-inhdr_error:
-       IP6_INC_STATS_BH(dev_net(dev), idev, IPSTATS_MIB_INHDRERRORS);
-drop:
-       return -1;
-}
-
-static void nf_bridge_update_protocol(struct sk_buff *skb)
-{
-       switch (skb->nf_bridge->orig_proto) {
-       case BRNF_PROTO_8021Q:
-               skb->protocol = htons(ETH_P_8021Q);
-               break;
-       case BRNF_PROTO_PPPOE:
-               skb->protocol = htons(ETH_P_PPP_SES);
-               break;
-       case BRNF_PROTO_UNCHANGED:
-               break;
-       }
-}
-
-/* Obtain the correct destination MAC address, while preserving the original
- * source MAC address. If we already know this address, we just copy it. If we
- * don't, we use the neighbour framework to find out. In both cases, we make
- * sure that br_handle_frame_finish() is called afterwards.
- */
-static int br_nf_pre_routing_finish_bridge(struct sock *sk, struct sk_buff *skb)
-{
-       struct neighbour *neigh;
-       struct dst_entry *dst;
-
-       skb->dev = bridge_parent(skb->dev);
-       if (!skb->dev)
-               goto free_skb;
-       dst = skb_dst(skb);
-       neigh = dst_neigh_lookup_skb(dst, skb);
-       if (neigh) {
-               struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
-               int ret;
-
-               if (neigh->hh.hh_len) {
-                       neigh_hh_bridge(&neigh->hh, skb);
-                       skb->dev = nf_bridge->physindev;
-                       ret = br_handle_frame_finish(sk, skb);
-               } else {
-                       /* the neighbour function below overwrites the complete
-                        * MAC header, so we save the Ethernet source address and
-                        * protocol number.
-                        */
-                       skb_copy_from_linear_data_offset(skb,
-                                                        -(ETH_HLEN-ETH_ALEN),
-                                                        nf_bridge->neigh_header,
-                                                        ETH_HLEN-ETH_ALEN);
-                       /* tell br_dev_xmit to continue with forwarding */
-                       nf_bridge->mask |= BRNF_BRIDGED_DNAT;
-                       /* FIXME Need to refragment */
-                       ret = neigh->output(neigh, skb);
-               }
-               neigh_release(neigh);
-               return ret;
-       }
-free_skb:
-       kfree_skb(skb);
-       return 0;
-}
-
-static bool daddr_was_changed(const struct sk_buff *skb,
-                             const struct nf_bridge_info *nf_bridge)
-{
-       switch (skb->protocol) {
-       case htons(ETH_P_IP):
-               return ip_hdr(skb)->daddr != nf_bridge->ipv4_daddr;
-       case htons(ETH_P_IPV6):
-               return memcmp(&nf_bridge->ipv6_daddr, &ipv6_hdr(skb)->daddr,
-                             sizeof(ipv6_hdr(skb)->daddr)) != 0;
-       default:
-               return false;
-       }
-}
-
-/* PF_BRIDGE/PRE_ROUTING: Undo the changes made for ip6tables
- * PREROUTING and continue the bridge PRE_ROUTING hook. See comment
- * for br_nf_pre_routing_finish(), same logic is used here but
- * equivalent IPv6 function ip6_route_input() called indirectly.
- */
-static int br_nf_pre_routing_finish_ipv6(struct sock *sk, struct sk_buff *skb)
-{
-       struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
-       struct rtable *rt;
-       struct net_device *dev = skb->dev;
-       const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops();
-
-       nf_bridge->frag_max_size = IP6CB(skb)->frag_max_size;
-
-       if (nf_bridge->pkt_otherhost) {
-               skb->pkt_type = PACKET_OTHERHOST;
-               nf_bridge->pkt_otherhost = false;
-       }
-       nf_bridge->mask &= ~BRNF_NF_BRIDGE_PREROUTING;
-       if (daddr_was_changed(skb, nf_bridge)) {
-               skb_dst_drop(skb);
-               v6ops->route_input(skb);
-
-               if (skb_dst(skb)->error) {
-                       kfree_skb(skb);
-                       return 0;
-               }
-
-               if (skb_dst(skb)->dev == dev) {
-                       skb->dev = nf_bridge->physindev;
-                       nf_bridge_update_protocol(skb);
-                       nf_bridge_push_encap_header(skb);
-                       NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING,
-                                      sk, skb, skb->dev, NULL,
-                                      br_nf_pre_routing_finish_bridge,
-                                      1);
-                       return 0;
-               }
-               ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr);
-               skb->pkt_type = PACKET_HOST;
-       } else {
-               rt = bridge_parent_rtable(nf_bridge->physindev);
-               if (!rt) {
-                       kfree_skb(skb);
-                       return 0;
-               }
-               skb_dst_set_noref(skb, &rt->dst);
-       }
-
-       skb->dev = nf_bridge->physindev;
-       nf_bridge_update_protocol(skb);
-       nf_bridge_push_encap_header(skb);
-       NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, sk, skb,
-                      skb->dev, NULL,
-                      br_handle_frame_finish, 1);
-
-       return 0;
-}
-
-/* This requires some explaining. If DNAT has taken place,
- * we will need to fix up the destination Ethernet address.
- * This is also true when SNAT takes place (for the reply direction).
- *
- * There are two cases to consider:
- * 1. The packet was DNAT'ed to a device in the same bridge
- *    port group as it was received on. We can still bridge
- *    the packet.
- * 2. The packet was DNAT'ed to a different device, either
- *    a non-bridged device or another bridge port group.
- *    The packet will need to be routed.
- *
- * The correct way of distinguishing between these two cases is to
- * call ip_route_input() and to look at skb->dst->dev, which is
- * changed to the destination device if ip_route_input() succeeds.
- *
- * Let's first consider the case that ip_route_input() succeeds:
- *
- * If the output device equals the logical bridge device the packet
- * came in on, we can consider this bridging. The corresponding MAC
- * address will be obtained in br_nf_pre_routing_finish_bridge.
- * Otherwise, the packet is considered to be routed and we just
- * change the destination MAC address so that the packet will
- * later be passed up to the IP stack to be routed. For a redirected
- * packet, ip_route_input() will give back the localhost as output device,
- * which differs from the bridge device.
- *
- * Let's now consider the case that ip_route_input() fails:
- *
- * This can be because the destination address is martian, in which case
- * the packet will be dropped.
- * If IP forwarding is disabled, ip_route_input() will fail, while
- * ip_route_output_key() can return success. The source
- * address for ip_route_output_key() is set to zero, so ip_route_output_key()
- * thinks we're handling a locally generated packet and won't care
- * if IP forwarding is enabled. If the output device equals the logical bridge
- * device, we proceed as if ip_route_input() succeeded. If it differs from the
- * logical bridge port or if ip_route_output_key() fails we drop the packet.
- */
-static int br_nf_pre_routing_finish(struct sock *sk, struct sk_buff *skb)
-{
-       struct net_device *dev = skb->dev;
-       struct iphdr *iph = ip_hdr(skb);
-       struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
-       struct rtable *rt;
-       int err;
-
-       nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
-
-       if (nf_bridge->pkt_otherhost) {
-               skb->pkt_type = PACKET_OTHERHOST;
-               nf_bridge->pkt_otherhost = false;
-       }
-       nf_bridge->mask &= ~BRNF_NF_BRIDGE_PREROUTING;
-       if (daddr_was_changed(skb, nf_bridge)) {
-               if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
-                       struct in_device *in_dev = __in_dev_get_rcu(dev);
-
-                       /* If err equals -EHOSTUNREACH the error is due to a
-                        * martian destination or due to the fact that
-                        * forwarding is disabled. For most martian packets,
-                        * ip_route_output_key() will fail. It won't fail for 2 types of
-                        * martian destinations: loopback destinations and destination
-                        * 0.0.0.0. In both cases the packet will be dropped because the
-                        * destination is the loopback device and not the bridge. */
-                       if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
-                               goto free_skb;
-
-                       rt = ip_route_output(dev_net(dev), iph->daddr, 0,
-                                            RT_TOS(iph->tos), 0);
-                       if (!IS_ERR(rt)) {
-                               /* - Bridged-and-DNAT'ed traffic doesn't
-                                *   require ip_forwarding. */
-                               if (rt->dst.dev == dev) {
-                                       skb_dst_set(skb, &rt->dst);
-                                       goto bridged_dnat;
-                               }
-                               ip_rt_put(rt);
-                       }
-free_skb:
-                       kfree_skb(skb);
-                       return 0;
-               } else {
-                       if (skb_dst(skb)->dev == dev) {
-bridged_dnat:
-                               skb->dev = nf_bridge->physindev;
-                               nf_bridge_update_protocol(skb);
-                               nf_bridge_push_encap_header(skb);
-                               NF_HOOK_THRESH(NFPROTO_BRIDGE,
-                                              NF_BR_PRE_ROUTING,
-                                              sk, skb, skb->dev, NULL,
-                                              br_nf_pre_routing_finish_bridge,
-                                              1);
-                               return 0;
-                       }
-                       ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr);
-                       skb->pkt_type = PACKET_HOST;
-               }
-       } else {
-               rt = bridge_parent_rtable(nf_bridge->physindev);
-               if (!rt) {
-                       kfree_skb(skb);
-                       return 0;
-               }
-               skb_dst_set_noref(skb, &rt->dst);
-       }
-
-       skb->dev = nf_bridge->physindev;
-       nf_bridge_update_protocol(skb);
-       nf_bridge_push_encap_header(skb);
-       NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, sk, skb,
-                      skb->dev, NULL,
-                      br_handle_frame_finish, 1);
-
-       return 0;
-}
-
-static struct net_device *brnf_get_logical_dev(struct sk_buff *skb, const struct net_device *dev)
-{
-       struct net_device *vlan, *br;
-
-       br = bridge_parent(dev);
-       if (brnf_pass_vlan_indev == 0 || !skb_vlan_tag_present(skb))
-               return br;
-
-       vlan = __vlan_find_dev_deep_rcu(br, skb->vlan_proto,
-                                   skb_vlan_tag_get(skb) & VLAN_VID_MASK);
-
-       return vlan ? vlan : br;
-}
-
-/* Some common code for IPv4/IPv6 */
-static struct net_device *setup_pre_routing(struct sk_buff *skb)
-{
-       struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
-
-       if (skb->pkt_type == PACKET_OTHERHOST) {
-               skb->pkt_type = PACKET_HOST;
-               nf_bridge->pkt_otherhost = true;
-       }
-
-       nf_bridge->mask |= BRNF_NF_BRIDGE_PREROUTING;
-       nf_bridge->physindev = skb->dev;
-       skb->dev = brnf_get_logical_dev(skb, skb->dev);
-
-       if (skb->protocol == htons(ETH_P_8021Q))
-               nf_bridge->orig_proto = BRNF_PROTO_8021Q;
-       else if (skb->protocol == htons(ETH_P_PPP_SES))
-               nf_bridge->orig_proto = BRNF_PROTO_PPPOE;
-
-       /* Must drop socket now because of tproxy. */
-       skb_orphan(skb);
-       return skb->dev;
-}
-
-/* Replicate the checks that IPv6 does on packet reception and pass the packet
- * to ip6tables.
- */
-static unsigned int br_nf_pre_routing_ipv6(const struct nf_hook_ops *ops,
-                                          struct sk_buff *skb,
-                                          const struct nf_hook_state *state)
-{
-       struct nf_bridge_info *nf_bridge;
-
-       if (br_validate_ipv6(skb))
-               return NF_DROP;
-
-       nf_bridge_put(skb->nf_bridge);
-       if (!nf_bridge_alloc(skb))
-               return NF_DROP;
-       if (!setup_pre_routing(skb))
-               return NF_DROP;
-
-       nf_bridge = nf_bridge_info_get(skb);
-       nf_bridge->ipv6_daddr = ipv6_hdr(skb)->daddr;
-
-       skb->protocol = htons(ETH_P_IPV6);
-       NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, state->sk, skb,
-               skb->dev, NULL,
-               br_nf_pre_routing_finish_ipv6);
-
-       return NF_STOLEN;
-}
-
-/* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
- * Replicate the checks that IPv4 does on packet reception.
- * Set skb->dev to the bridge device (i.e. parent of the
- * receiving device) to make netfilter happy, the REDIRECT
- * target in particular.  Save the original destination IP
- * address to be able to detect DNAT afterwards. */
-static unsigned int br_nf_pre_routing(const struct nf_hook_ops *ops,
-                                     struct sk_buff *skb,
-                                     const struct nf_hook_state *state)
-{
-       struct nf_bridge_info *nf_bridge;
-       struct net_bridge_port *p;
-       struct net_bridge *br;
-       __u32 len = nf_bridge_encap_header_len(skb);
-
-       if (unlikely(!pskb_may_pull(skb, len)))
-               return NF_DROP;
-
-       p = br_port_get_rcu(state->in);
-       if (p == NULL)
-               return NF_DROP;
-       br = p->br;
-
-       if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb)) {
-               if (!brnf_call_ip6tables && !br->nf_call_ip6tables)
-                       return NF_ACCEPT;
-
-               nf_bridge_pull_encap_header_rcsum(skb);
-               return br_nf_pre_routing_ipv6(ops, skb, state);
-       }
-
-       if (!brnf_call_iptables && !br->nf_call_iptables)
-               return NF_ACCEPT;
-
-       if (!IS_IP(skb) && !IS_VLAN_IP(skb) && !IS_PPPOE_IP(skb))
-               return NF_ACCEPT;
-
-       nf_bridge_pull_encap_header_rcsum(skb);
-
-       if (br_validate_ipv4(skb))
-               return NF_DROP;
-
-       nf_bridge_put(skb->nf_bridge);
-       if (!nf_bridge_alloc(skb))
-               return NF_DROP;
-       if (!setup_pre_routing(skb))
-               return NF_DROP;
-
-       nf_bridge = nf_bridge_info_get(skb);
-       nf_bridge->ipv4_daddr = ip_hdr(skb)->daddr;
-
-       skb->protocol = htons(ETH_P_IP);
-
-       NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, state->sk, skb,
-               skb->dev, NULL,
-               br_nf_pre_routing_finish);
-
-       return NF_STOLEN;
-}
-
-
-/* PF_BRIDGE/LOCAL_IN ************************************************/
-/* The packet is locally destined, which requires a real
- * dst_entry, so detach the fake one.  On the way up, the
- * packet would pass through PRE_ROUTING again (which already
- * took place when the packet entered the bridge), but we
- * register an IPv4 PRE_ROUTING 'sabotage' hook that will
- * prevent this from happening. */
-static unsigned int br_nf_local_in(const struct nf_hook_ops *ops,
-                                  struct sk_buff *skb,
-                                  const struct nf_hook_state *state)
-{
-       br_drop_fake_rtable(skb);
-       return NF_ACCEPT;
-}
-
-/* PF_BRIDGE/FORWARD *************************************************/
-static int br_nf_forward_finish(struct sock *sk, struct sk_buff *skb)
-{
-       struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
-       struct net_device *in;
-
-       if (!IS_ARP(skb) && !IS_VLAN_ARP(skb)) {
-
-               if (skb->protocol == htons(ETH_P_IP))
-                       nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
-
-               if (skb->protocol == htons(ETH_P_IPV6))
-                       nf_bridge->frag_max_size = IP6CB(skb)->frag_max_size;
-
-               in = nf_bridge->physindev;
-               if (nf_bridge->pkt_otherhost) {
-                       skb->pkt_type = PACKET_OTHERHOST;
-                       nf_bridge->pkt_otherhost = false;
-               }
-               nf_bridge_update_protocol(skb);
-       } else {
-               in = *((struct net_device **)(skb->cb));
-       }
-       nf_bridge_push_encap_header(skb);
-
-       NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_FORWARD, sk, skb,
-                      in, skb->dev, br_forward_finish, 1);
-       return 0;
-}
-
-
-/* This is the 'purely bridged' case.  For IP, we pass the packet to
- * netfilter with indev and outdev set to the bridge device,
- * but we are still able to filter on the 'real' indev/outdev
- * because of the physdev module. For ARP, indev and outdev are the
- * bridge ports. */
-static unsigned int br_nf_forward_ip(const struct nf_hook_ops *ops,
-                                    struct sk_buff *skb,
-                                    const struct nf_hook_state *state)
-{
-       struct nf_bridge_info *nf_bridge;
-       struct net_device *parent;
-       u_int8_t pf;
-
-       if (!skb->nf_bridge)
-               return NF_ACCEPT;
-
-       /* Need exclusive nf_bridge_info since we might have multiple
-        * different physoutdevs. */
-       if (!nf_bridge_unshare(skb))
-               return NF_DROP;
-
-       nf_bridge = nf_bridge_info_get(skb);
-       if (!nf_bridge)
-               return NF_DROP;
-
-       parent = bridge_parent(state->out);
-       if (!parent)
-               return NF_DROP;
-
-       if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
-               pf = NFPROTO_IPV4;
-       else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
-               pf = NFPROTO_IPV6;
-       else
-               return NF_ACCEPT;
-
-       nf_bridge_pull_encap_header(skb);
-
-       if (skb->pkt_type == PACKET_OTHERHOST) {
-               skb->pkt_type = PACKET_HOST;
-               nf_bridge->pkt_otherhost = true;
-       }
-
-       if (pf == NFPROTO_IPV4) {
-               if (br_validate_ipv4(skb))
-                       return NF_DROP;
-               IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
-       }
-
-       if (pf == NFPROTO_IPV6) {
-               if (br_validate_ipv6(skb))
-                       return NF_DROP;
-               IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
-       }
-
-       nf_bridge->physoutdev = skb->dev;
-       if (pf == NFPROTO_IPV4)
-               skb->protocol = htons(ETH_P_IP);
-       else
-               skb->protocol = htons(ETH_P_IPV6);
-
-       NF_HOOK(pf, NF_INET_FORWARD, NULL, skb,
-               brnf_get_logical_dev(skb, state->in),
-               parent, br_nf_forward_finish);
-
-       return NF_STOLEN;
-}
-
-static unsigned int br_nf_forward_arp(const struct nf_hook_ops *ops,
-                                     struct sk_buff *skb,
-                                     const struct nf_hook_state *state)
-{
-       struct net_bridge_port *p;
-       struct net_bridge *br;
-       struct net_device **d = (struct net_device **)(skb->cb);
-
-       p = br_port_get_rcu(state->out);
-       if (p == NULL)
-               return NF_ACCEPT;
-       br = p->br;
-
-       if (!brnf_call_arptables && !br->nf_call_arptables)
-               return NF_ACCEPT;
-
-       if (!IS_ARP(skb)) {
-               if (!IS_VLAN_ARP(skb))
-                       return NF_ACCEPT;
-               nf_bridge_pull_encap_header(skb);
-       }
-
-       if (arp_hdr(skb)->ar_pln != 4) {
-               if (IS_VLAN_ARP(skb))
-                       nf_bridge_push_encap_header(skb);
-               return NF_ACCEPT;
-       }
-       *d = state->in;
-       NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, state->sk, skb,
-               state->in, state->out, br_nf_forward_finish);
-
-       return NF_STOLEN;
-}
-
-#if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4) || IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
-static int br_nf_push_frag_xmit(struct sock *sk, struct sk_buff *skb)
-{
-       struct brnf_frag_data *data;
-       int err;
-
-       data = this_cpu_ptr(&brnf_frag_data_storage);
-       err = skb_cow_head(skb, data->size);
-
-       if (err) {
-               kfree_skb(skb);
-               return 0;
-       }
-
-       if (data->vlan_tci) {
-               skb->vlan_tci = data->vlan_tci;
-               skb->vlan_proto = data->vlan_proto;
-       }
-
-       skb_copy_to_linear_data_offset(skb, -data->size, data->mac, data->size);
-       __skb_push(skb, data->encap_size);
-
-       nf_bridge_info_free(skb);
-       return br_dev_queue_push_xmit(sk, skb);
-}
-#endif
-
-static int br_nf_ip_fragment(struct sock *sk, struct sk_buff *skb,
-                            int (*output)(struct sock *, struct sk_buff *))
-{
-       unsigned int mtu = ip_skb_dst_mtu(skb);
-       struct iphdr *iph = ip_hdr(skb);
-       struct rtable *rt = skb_rtable(skb);
-       struct net_device *dev = rt->dst.dev;
-
-       if (unlikely(((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) ||
-                    (IPCB(skb)->frag_max_size &&
-                     IPCB(skb)->frag_max_size > mtu))) {
-               IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
-               kfree_skb(skb);
-               return -EMSGSIZE;
-       }
-
-       return ip_do_fragment(sk, skb, output);
-}
-
-static unsigned int nf_bridge_mtu_reduction(const struct sk_buff *skb)
-{
-       if (skb->nf_bridge->orig_proto == BRNF_PROTO_PPPOE)
-               return PPPOE_SES_HLEN;
-       return 0;
-}
-
-static int br_nf_dev_queue_xmit(struct sock *sk, struct sk_buff *skb)
-{
-       struct nf_bridge_info *nf_bridge;
-       unsigned int mtu_reserved;
-
-       mtu_reserved = nf_bridge_mtu_reduction(skb);
-
-       if (skb_is_gso(skb) || skb->len + mtu_reserved <= skb->dev->mtu) {
-               nf_bridge_info_free(skb);
-               return br_dev_queue_push_xmit(sk, skb);
-       }
-
-       nf_bridge = nf_bridge_info_get(skb);
-
-#if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4)
-       /* This is wrong! We should preserve the original fragment
-        * boundaries by preserving frag_list rather than refragmenting.
-        */
-       if (skb->protocol == htons(ETH_P_IP)) {
-               struct brnf_frag_data *data;
-
-               if (br_validate_ipv4(skb))
-                       return NF_DROP;
-
-               IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
-
-               nf_bridge_update_protocol(skb);
-
-               data = this_cpu_ptr(&brnf_frag_data_storage);
-
-               data->vlan_tci = skb->vlan_tci;
-               data->vlan_proto = skb->vlan_proto;
-               data->encap_size = nf_bridge_encap_header_len(skb);
-               data->size = ETH_HLEN + data->encap_size;
-
-               skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
-                                                data->size);
-
-               return br_nf_ip_fragment(sk, skb, br_nf_push_frag_xmit);
-       }
-#endif
-#if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
-       if (skb->protocol == htons(ETH_P_IPV6)) {
-               const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops();
-               struct brnf_frag_data *data;
-
-               if (br_validate_ipv6(skb))
-                       return NF_DROP;
-
-               IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
-
-               nf_bridge_update_protocol(skb);
-
-               data = this_cpu_ptr(&brnf_frag_data_storage);
-               data->encap_size = nf_bridge_encap_header_len(skb);
-               data->size = ETH_HLEN + data->encap_size;
-
-               skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
-                                                data->size);
-
-               if (v6ops)
-                       return v6ops->fragment(sk, skb, br_nf_push_frag_xmit);
-               else
-                       return -EMSGSIZE;
-       }
-#endif
-       nf_bridge_info_free(skb);
-       return br_dev_queue_push_xmit(sk, skb);
-}
-
-/* PF_BRIDGE/POST_ROUTING ********************************************/
-static unsigned int br_nf_post_routing(const struct nf_hook_ops *ops,
-                                      struct sk_buff *skb,
-                                      const struct nf_hook_state *state)
-{
-       struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
-       struct net_device *realoutdev = bridge_parent(skb->dev);
-       u_int8_t pf;
-
-       /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
-        * on a bridge, but was delivered locally and is now being routed:
-        *
-        * POST_ROUTING was already invoked from the ip stack.
-        */
-       if (!nf_bridge || !nf_bridge->physoutdev)
-               return NF_ACCEPT;
-
-       if (!realoutdev)
-               return NF_DROP;
-
-       if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
-               pf = NFPROTO_IPV4;
-       else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
-               pf = NFPROTO_IPV6;
-       else
-               return NF_ACCEPT;
-
-       /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
-        * about the value of skb->pkt_type. */
-       if (skb->pkt_type == PACKET_OTHERHOST) {
-               skb->pkt_type = PACKET_HOST;
-               nf_bridge->pkt_otherhost = true;
-       }
-
-       nf_bridge_pull_encap_header(skb);
-       if (pf == NFPROTO_IPV4)
-               skb->protocol = htons(ETH_P_IP);
-       else
-               skb->protocol = htons(ETH_P_IPV6);
-
-       NF_HOOK(pf, NF_INET_POST_ROUTING, state->sk, skb,
-               NULL, realoutdev,
-               br_nf_dev_queue_xmit);
-
-       return NF_STOLEN;
-}
-
-/* IP/SABOTAGE *****************************************************/
-/* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
- * for the second time. */
-static unsigned int ip_sabotage_in(const struct nf_hook_ops *ops,
-                                  struct sk_buff *skb,
-                                  const struct nf_hook_state *state)
-{
-       if (skb->nf_bridge &&
-           !(skb->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)) {
-               return NF_STOP;
-       }
-
-       return NF_ACCEPT;
-}
-
-/* This is called when br_netfilter has called into iptables/netfilter,
- * and DNAT has taken place on a bridge-forwarded packet.
- *
- * neigh->output has created a new MAC header, with local br0 MAC
- * as saddr.
- *
- * This restores the original MAC saddr of the bridged packet
- * before invoking bridge forward logic to transmit the packet.
- */
-static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff *skb)
-{
-       struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
-
-       skb_pull(skb, ETH_HLEN);
-       nf_bridge->mask &= ~BRNF_BRIDGED_DNAT;
-
-       BUILD_BUG_ON(sizeof(nf_bridge->neigh_header) != (ETH_HLEN - ETH_ALEN));
-
-       skb_copy_to_linear_data_offset(skb, -(ETH_HLEN - ETH_ALEN),
-                                      nf_bridge->neigh_header,
-                                      ETH_HLEN - ETH_ALEN);
-       skb->dev = nf_bridge->physindev;
-
-       nf_bridge->physoutdev = NULL;
-       br_handle_frame_finish(NULL, skb);
-}
-
-static int br_nf_dev_xmit(struct sk_buff *skb)
-{
-       if (skb->nf_bridge && (skb->nf_bridge->mask & BRNF_BRIDGED_DNAT)) {
-               br_nf_pre_routing_finish_bridge_slow(skb);
-               return 1;
-       }
-       return 0;
-}
-
-static const struct nf_br_ops br_ops = {
-       .br_dev_xmit_hook =     br_nf_dev_xmit,
-};
-
-void br_netfilter_enable(void)
-{
-}
-EXPORT_SYMBOL_GPL(br_netfilter_enable);
-
-/* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
- * br_dev_queue_push_xmit is called afterwards */
-static struct nf_hook_ops br_nf_ops[] __read_mostly = {
-       {
-               .hook = br_nf_pre_routing,
-               .owner = THIS_MODULE,
-               .pf = NFPROTO_BRIDGE,
-               .hooknum = NF_BR_PRE_ROUTING,
-               .priority = NF_BR_PRI_BRNF,
-       },
-       {
-               .hook = br_nf_local_in,
-               .owner = THIS_MODULE,
-               .pf = NFPROTO_BRIDGE,
-               .hooknum = NF_BR_LOCAL_IN,
-               .priority = NF_BR_PRI_BRNF,
-       },
-       {
-               .hook = br_nf_forward_ip,
-               .owner = THIS_MODULE,
-               .pf = NFPROTO_BRIDGE,
-               .hooknum = NF_BR_FORWARD,
-               .priority = NF_BR_PRI_BRNF - 1,
-       },
-       {
-               .hook = br_nf_forward_arp,
-               .owner = THIS_MODULE,
-               .pf = NFPROTO_BRIDGE,
-               .hooknum = NF_BR_FORWARD,
-               .priority = NF_BR_PRI_BRNF,
-       },
-       {
-               .hook = br_nf_post_routing,
-               .owner = THIS_MODULE,
-               .pf = NFPROTO_BRIDGE,
-               .hooknum = NF_BR_POST_ROUTING,
-               .priority = NF_BR_PRI_LAST,
-       },
-       {
-               .hook = ip_sabotage_in,
-               .owner = THIS_MODULE,
-               .pf = NFPROTO_IPV4,
-               .hooknum = NF_INET_PRE_ROUTING,
-               .priority = NF_IP_PRI_FIRST,
-       },
-       {
-               .hook = ip_sabotage_in,
-               .owner = THIS_MODULE,
-               .pf = NFPROTO_IPV6,
-               .hooknum = NF_INET_PRE_ROUTING,
-               .priority = NF_IP6_PRI_FIRST,
-       },
-};
-
-#ifdef CONFIG_SYSCTL
-static
-int brnf_sysctl_call_tables(struct ctl_table *ctl, int write,
-                           void __user *buffer, size_t *lenp, loff_t *ppos)
-{
-       int ret;
-
-       ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
-
-       if (write && *(int *)(ctl->data))
-               *(int *)(ctl->data) = 1;
-       return ret;
-}
-
-static struct ctl_table brnf_table[] = {
-       {
-               .procname       = "bridge-nf-call-arptables",
-               .data           = &brnf_call_arptables,
-               .maxlen         = sizeof(int),
-               .mode           = 0644,
-               .proc_handler   = brnf_sysctl_call_tables,
-       },
-       {
-               .procname       = "bridge-nf-call-iptables",
-               .data           = &brnf_call_iptables,
-               .maxlen         = sizeof(int),
-               .mode           = 0644,
-               .proc_handler   = brnf_sysctl_call_tables,
-       },
-       {
-               .procname       = "bridge-nf-call-ip6tables",
-               .data           = &brnf_call_ip6tables,
-               .maxlen         = sizeof(int),
-               .mode           = 0644,
-               .proc_handler   = brnf_sysctl_call_tables,
-       },
-       {
-               .procname       = "bridge-nf-filter-vlan-tagged",
-               .data           = &brnf_filter_vlan_tagged,
-               .maxlen         = sizeof(int),
-               .mode           = 0644,
-               .proc_handler   = brnf_sysctl_call_tables,
-       },
-       {
-               .procname       = "bridge-nf-filter-pppoe-tagged",
-               .data           = &brnf_filter_pppoe_tagged,
-               .maxlen         = sizeof(int),
-               .mode           = 0644,
-               .proc_handler   = brnf_sysctl_call_tables,
-       },
-       {
-               .procname       = "bridge-nf-pass-vlan-input-dev",
-               .data           = &brnf_pass_vlan_indev,
-               .maxlen         = sizeof(int),
-               .mode           = 0644,
-               .proc_handler   = brnf_sysctl_call_tables,
-       },
-       { }
-};
-#endif
-
-static int __init br_netfilter_init(void)
-{
-       int ret;
-
-       ret = nf_register_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
-       if (ret < 0)
-               return ret;
-
-#ifdef CONFIG_SYSCTL
-       brnf_sysctl_header = register_net_sysctl(&init_net, "net/bridge", brnf_table);
-       if (brnf_sysctl_header == NULL) {
-               printk(KERN_WARNING
-                      "br_netfilter: can't register to sysctl.\n");
-               nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
-               return -ENOMEM;
-       }
-#endif
-       RCU_INIT_POINTER(nf_br_ops, &br_ops);
-       printk(KERN_NOTICE "Bridge firewalling registered\n");
-       return 0;
-}
-
-static void __exit br_netfilter_fini(void)
-{
-       RCU_INIT_POINTER(nf_br_ops, NULL);
-       nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
-#ifdef CONFIG_SYSCTL
-       unregister_net_sysctl_table(brnf_sysctl_header);
-#endif
-}
-
-module_init(br_netfilter_init);
-module_exit(br_netfilter_fini);
-
-MODULE_LICENSE("GPL");
-MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
-MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
-MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");
diff --git a/net/bridge/br_netfilter_hooks.c b/net/bridge/br_netfilter_hooks.c
new file mode 100644 (file)
index 0000000..e4e5f2f
--- /dev/null
@@ -0,0 +1,1282 @@
+/*
+ *     Handle firewalling
+ *     Linux ethernet bridge
+ *
+ *     Authors:
+ *     Lennert Buytenhek               <buytenh@gnu.org>
+ *     Bart De Schuymer                <bdschuym@pandora.be>
+ *
+ *     This program is free software; you can redistribute it and/or
+ *     modify it under the terms of the GNU General Public License
+ *     as published by the Free Software Foundation; either version
+ *     2 of the License, or (at your option) any later version.
+ *
+ *     Lennert dedicates this file to Kerstin Wurdinger.
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/ip.h>
+#include <linux/netdevice.h>
+#include <linux/skbuff.h>
+#include <linux/if_arp.h>
+#include <linux/if_ether.h>
+#include <linux/if_vlan.h>
+#include <linux/if_pppox.h>
+#include <linux/ppp_defs.h>
+#include <linux/netfilter_bridge.h>
+#include <linux/netfilter_ipv4.h>
+#include <linux/netfilter_ipv6.h>
+#include <linux/netfilter_arp.h>
+#include <linux/in_route.h>
+#include <linux/inetdevice.h>
+
+#include <net/ip.h>
+#include <net/ipv6.h>
+#include <net/addrconf.h>
+#include <net/route.h>
+#include <net/netfilter/br_netfilter.h>
+
+#include <asm/uaccess.h>
+#include "br_private.h"
+#ifdef CONFIG_SYSCTL
+#include <linux/sysctl.h>
+#endif
+
+#ifdef CONFIG_SYSCTL
+static struct ctl_table_header *brnf_sysctl_header;
+static int brnf_call_iptables __read_mostly = 1;
+static int brnf_call_ip6tables __read_mostly = 1;
+static int brnf_call_arptables __read_mostly = 1;
+static int brnf_filter_vlan_tagged __read_mostly = 0;
+static int brnf_filter_pppoe_tagged __read_mostly = 0;
+static int brnf_pass_vlan_indev __read_mostly = 0;
+#else
+#define brnf_call_iptables 1
+#define brnf_call_ip6tables 1
+#define brnf_call_arptables 1
+#define brnf_filter_vlan_tagged 0
+#define brnf_filter_pppoe_tagged 0
+#define brnf_pass_vlan_indev 0
+#endif
+
+#define IS_IP(skb) \
+       (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP))
+
+#define IS_IPV6(skb) \
+       (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6))
+
+#define IS_ARP(skb) \
+       (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP))
+
+static inline __be16 vlan_proto(const struct sk_buff *skb)
+{
+       if (skb_vlan_tag_present(skb))
+               return skb->protocol;
+       else if (skb->protocol == htons(ETH_P_8021Q))
+               return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
+       else
+               return 0;
+}
+
+#define IS_VLAN_IP(skb) \
+       (vlan_proto(skb) == htons(ETH_P_IP) && \
+        brnf_filter_vlan_tagged)
+
+#define IS_VLAN_IPV6(skb) \
+       (vlan_proto(skb) == htons(ETH_P_IPV6) && \
+        brnf_filter_vlan_tagged)
+
+#define IS_VLAN_ARP(skb) \
+       (vlan_proto(skb) == htons(ETH_P_ARP) && \
+        brnf_filter_vlan_tagged)
+
+static inline __be16 pppoe_proto(const struct sk_buff *skb)
+{
+       return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
+                           sizeof(struct pppoe_hdr)));
+}
+
+#define IS_PPPOE_IP(skb) \
+       (skb->protocol == htons(ETH_P_PPP_SES) && \
+        pppoe_proto(skb) == htons(PPP_IP) && \
+        brnf_filter_pppoe_tagged)
+
+#define IS_PPPOE_IPV6(skb) \
+       (skb->protocol == htons(ETH_P_PPP_SES) && \
+        pppoe_proto(skb) == htons(PPP_IPV6) && \
+        brnf_filter_pppoe_tagged)
+
+/* largest possible L2 header, see br_nf_dev_queue_xmit() */
+#define NF_BRIDGE_MAX_MAC_HEADER_LENGTH (PPPOE_SES_HLEN + ETH_HLEN)
+
+#if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4)
+struct brnf_frag_data {
+       char mac[NF_BRIDGE_MAX_MAC_HEADER_LENGTH];
+       u8 encap_size;
+       u8 size;
+       u16 vlan_tci;
+       __be16 vlan_proto;
+};
+
+static DEFINE_PER_CPU(struct brnf_frag_data, brnf_frag_data_storage);
+#endif
+
+static struct nf_bridge_info *nf_bridge_info_get(const struct sk_buff *skb)
+{
+       return skb->nf_bridge;
+}
+
+static void nf_bridge_info_free(struct sk_buff *skb)
+{
+       if (skb->nf_bridge) {
+               nf_bridge_put(skb->nf_bridge);
+               skb->nf_bridge = NULL;
+       }
+}
+
+static inline struct rtable *bridge_parent_rtable(const struct net_device *dev)
+{
+       struct net_bridge_port *port;
+
+       port = br_port_get_rcu(dev);
+       return port ? &port->br->fake_rtable : NULL;
+}
+
+static inline struct net_device *bridge_parent(const struct net_device *dev)
+{
+       struct net_bridge_port *port;
+
+       port = br_port_get_rcu(dev);
+       return port ? port->br->dev : NULL;
+}
+
+static inline struct nf_bridge_info *nf_bridge_alloc(struct sk_buff *skb)
+{
+       skb->nf_bridge = kzalloc(sizeof(struct nf_bridge_info), GFP_ATOMIC);
+       if (likely(skb->nf_bridge))
+               atomic_set(&(skb->nf_bridge->use), 1);
+
+       return skb->nf_bridge;
+}
+
+static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
+{
+       struct nf_bridge_info *nf_bridge = skb->nf_bridge;
+
+       if (atomic_read(&nf_bridge->use) > 1) {
+               struct nf_bridge_info *tmp = nf_bridge_alloc(skb);
+
+               if (tmp) {
+                       memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info));
+                       atomic_set(&tmp->use, 1);
+               }
+               nf_bridge_put(nf_bridge);
+               nf_bridge = tmp;
+       }
+       return nf_bridge;
+}
+
+static unsigned int nf_bridge_encap_header_len(const struct sk_buff *skb)
+{
+       switch (skb->protocol) {
+       case __cpu_to_be16(ETH_P_8021Q):
+               return VLAN_HLEN;
+       case __cpu_to_be16(ETH_P_PPP_SES):
+               return PPPOE_SES_HLEN;
+       default:
+               return 0;
+       }
+}
+
+static inline void nf_bridge_push_encap_header(struct sk_buff *skb)
+{
+       unsigned int len = nf_bridge_encap_header_len(skb);
+
+       skb_push(skb, len);
+       skb->network_header -= len;
+}
+
+static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
+{
+       unsigned int len = nf_bridge_encap_header_len(skb);
+
+       skb_pull(skb, len);
+       skb->network_header += len;
+}
+
+static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
+{
+       unsigned int len = nf_bridge_encap_header_len(skb);
+
+       skb_pull_rcsum(skb, len);
+       skb->network_header += len;
+}
+
+/* When handing a packet over to the IP layer
+ * check whether we have a skb that is in the
+ * expected format
+ */
+
+static int br_validate_ipv4(struct sk_buff *skb)
+{
+       const struct iphdr *iph;
+       struct net_device *dev = skb->dev;
+       u32 len;
+
+       if (!pskb_may_pull(skb, sizeof(struct iphdr)))
+               goto inhdr_error;
+
+       iph = ip_hdr(skb);
+
+       /* Basic sanity checks */
+       if (iph->ihl < 5 || iph->version != 4)
+               goto inhdr_error;
+
+       if (!pskb_may_pull(skb, iph->ihl*4))
+               goto inhdr_error;
+
+       iph = ip_hdr(skb);
+       if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
+               goto inhdr_error;
+
+       len = ntohs(iph->tot_len);
+       if (skb->len < len) {
+               IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INTRUNCATEDPKTS);
+               goto drop;
+       } else if (len < (iph->ihl*4))
+               goto inhdr_error;
+
+       if (pskb_trim_rcsum(skb, len)) {
+               IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
+               goto drop;
+       }
+
+       memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
+       /* We should really parse IP options here but until
+        * somebody who actually uses IP options complains to
+        * us we'll just silently ignore the options because
+        * we're lazy!
+        */
+       return 0;
+
+inhdr_error:
+       IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
+drop:
+       return -1;
+}
+
+/* We only check the length. A bridge shouldn't do any hop-by-hop stuff
+ * anyway
+ */
+static int check_hbh_len(struct sk_buff *skb)
+{
+       unsigned char *raw = (u8 *)(ipv6_hdr(skb) + 1);
+       u32 pkt_len;
+       const unsigned char *nh = skb_network_header(skb);
+       int off = raw - nh;
+       int len = (raw[1] + 1) << 3;
+
+       if ((raw + len) - skb->data > skb_headlen(skb))
+               goto bad;
+
+       off += 2;
+       len -= 2;
+
+       while (len > 0) {
+               int optlen = nh[off + 1] + 2;
+
+               switch (nh[off]) {
+               case IPV6_TLV_PAD1:
+                       optlen = 1;
+                       break;
+
+               case IPV6_TLV_PADN:
+                       break;
+
+               case IPV6_TLV_JUMBO:
+                       if (nh[off + 1] != 4 || (off & 3) != 2)
+                               goto bad;
+                       pkt_len = ntohl(*(__be32 *)(nh + off + 2));
+                       if (pkt_len <= IPV6_MAXPLEN ||
+                           ipv6_hdr(skb)->payload_len)
+                               goto bad;
+                       if (pkt_len > skb->len - sizeof(struct ipv6hdr))
+                               goto bad;
+                       if (pskb_trim_rcsum(skb,
+                                           pkt_len + sizeof(struct ipv6hdr)))
+                               goto bad;
+                       nh = skb_network_header(skb);
+                       break;
+               default:
+                       if (optlen > len)
+                               goto bad;
+                       break;
+               }
+               off += optlen;
+               len -= optlen;
+       }
+       if (len == 0)
+               return 0;
+bad:
+       return -1;
+}
+
+/* Equivalent to br_validate_ipv4 for IPv6 */
+static int br_validate_ipv6(struct sk_buff *skb)
+{
+       const struct ipv6hdr *hdr;
+       struct net_device *dev = skb->dev;
+       struct inet6_dev *idev = in6_dev_get(skb->dev);
+       u32 pkt_len;
+       u8 ip6h_len = sizeof(struct ipv6hdr);
+
+       if (!pskb_may_pull(skb, ip6h_len))
+               goto inhdr_error;
+
+       if (skb->len < ip6h_len)
+               goto drop;
+
+       hdr = ipv6_hdr(skb);
+
+       if (hdr->version != 6)
+               goto inhdr_error;
+
+       pkt_len = ntohs(hdr->payload_len);
+
+       if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
+               if (pkt_len + ip6h_len > skb->len) {
+                       IP6_INC_STATS_BH(dev_net(dev), idev,
+                                        IPSTATS_MIB_INTRUNCATEDPKTS);
+                       goto drop;
+               }
+               if (pskb_trim_rcsum(skb, pkt_len + ip6h_len)) {
+                       IP6_INC_STATS_BH(dev_net(dev), idev,
+                                        IPSTATS_MIB_INDISCARDS);
+                       goto drop;
+               }
+       }
+       if (hdr->nexthdr == NEXTHDR_HOP && check_hbh_len(skb))
+               goto drop;
+
+       memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
+       /* No IP options in IPv6 header; however it should be
+        * checked if some next headers need special treatment
+        */
+       return 0;
+
+inhdr_error:
+       IP6_INC_STATS_BH(dev_net(dev), idev, IPSTATS_MIB_INHDRERRORS);
+drop:
+       return -1;
+}
+
+static void nf_bridge_update_protocol(struct sk_buff *skb)
+{
+       switch (skb->nf_bridge->orig_proto) {
+       case BRNF_PROTO_8021Q:
+               skb->protocol = htons(ETH_P_8021Q);
+               break;
+       case BRNF_PROTO_PPPOE:
+               skb->protocol = htons(ETH_P_PPP_SES);
+               break;
+       case BRNF_PROTO_UNCHANGED:
+               break;
+       }
+}
+
+/* Obtain the correct destination MAC address, while preserving the original
+ * source MAC address. If we already know this address, we just copy it. If we
+ * don't, we use the neighbour framework to find out. In both cases, we make
+ * sure that br_handle_frame_finish() is called afterwards.
+ */
+static int br_nf_pre_routing_finish_bridge(struct sock *sk, struct sk_buff *skb)
+{
+       struct neighbour *neigh;
+       struct dst_entry *dst;
+
+       skb->dev = bridge_parent(skb->dev);
+       if (!skb->dev)
+               goto free_skb;
+       dst = skb_dst(skb);
+       neigh = dst_neigh_lookup_skb(dst, skb);
+       if (neigh) {
+               struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
+               int ret;
+
+               if (neigh->hh.hh_len) {
+                       neigh_hh_bridge(&neigh->hh, skb);
+                       skb->dev = nf_bridge->physindev;
+                       ret = br_handle_frame_finish(sk, skb);
+               } else {
+                       /* the neighbour function below overwrites the complete
+                        * MAC header, so we save the Ethernet source address and
+                        * protocol number.
+                        */
+                       skb_copy_from_linear_data_offset(skb,
+                                                        -(ETH_HLEN-ETH_ALEN),
+                                                        nf_bridge->neigh_header,
+                                                        ETH_HLEN-ETH_ALEN);
+                       /* tell br_dev_xmit to continue with forwarding */
+                       nf_bridge->mask |= BRNF_BRIDGED_DNAT;
+                       /* FIXME Need to refragment */
+                       ret = neigh->output(neigh, skb);
+               }
+               neigh_release(neigh);
+               return ret;
+       }
+free_skb:
+       kfree_skb(skb);
+       return 0;
+}
+
+static bool daddr_was_changed(const struct sk_buff *skb,
+                             const struct nf_bridge_info *nf_bridge)
+{
+       switch (skb->protocol) {
+       case htons(ETH_P_IP):
+               return ip_hdr(skb)->daddr != nf_bridge->ipv4_daddr;
+       case htons(ETH_P_IPV6):
+               return memcmp(&nf_bridge->ipv6_daddr, &ipv6_hdr(skb)->daddr,
+                             sizeof(ipv6_hdr(skb)->daddr)) != 0;
+       default:
+               return false;
+       }
+}
+
+/* PF_BRIDGE/PRE_ROUTING: Undo the changes made for ip6tables
+ * PREROUTING and continue the bridge PRE_ROUTING hook. See comment
+ * for br_nf_pre_routing_finish(), same logic is used here but
+ * equivalent IPv6 function ip6_route_input() called indirectly.
+ */
+static int br_nf_pre_routing_finish_ipv6(struct sock *sk, struct sk_buff *skb)
+{
+       struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
+       struct rtable *rt;
+       struct net_device *dev = skb->dev;
+       const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops();
+
+       nf_bridge->frag_max_size = IP6CB(skb)->frag_max_size;
+
+       if (nf_bridge->pkt_otherhost) {
+               skb->pkt_type = PACKET_OTHERHOST;
+               nf_bridge->pkt_otherhost = false;
+       }
+       nf_bridge->mask &= ~BRNF_NF_BRIDGE_PREROUTING;
+       if (daddr_was_changed(skb, nf_bridge)) {
+               skb_dst_drop(skb);
+               v6ops->route_input(skb);
+
+               if (skb_dst(skb)->error) {
+                       kfree_skb(skb);
+                       return 0;
+               }
+
+               if (skb_dst(skb)->dev == dev) {
+                       skb->dev = nf_bridge->physindev;
+                       nf_bridge_update_protocol(skb);
+                       nf_bridge_push_encap_header(skb);
+                       NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING,
+                                      sk, skb, skb->dev, NULL,
+                                      br_nf_pre_routing_finish_bridge,
+                                      1);
+                       return 0;
+               }
+               ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr);
+               skb->pkt_type = PACKET_HOST;
+       } else {
+               rt = bridge_parent_rtable(nf_bridge->physindev);
+               if (!rt) {
+                       kfree_skb(skb);
+                       return 0;
+               }
+               skb_dst_set_noref(skb, &rt->dst);
+       }
+
+       skb->dev = nf_bridge->physindev;
+       nf_bridge_update_protocol(skb);
+       nf_bridge_push_encap_header(skb);
+       NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, sk, skb,
+                      skb->dev, NULL,
+                      br_handle_frame_finish, 1);
+
+       return 0;
+}
+
+/* This requires some explaining. If DNAT has taken place,
+ * we will need to fix up the destination Ethernet address.
+ * This is also true when SNAT takes place (for the reply direction).
+ *
+ * There are two cases to consider:
+ * 1. The packet was DNAT'ed to a device in the same bridge
+ *    port group as it was received on. We can still bridge
+ *    the packet.
+ * 2. The packet was DNAT'ed to a different device, either
+ *    a non-bridged device or another bridge port group.
+ *    The packet will need to be routed.
+ *
+ * The correct way of distinguishing between these two cases is to
+ * call ip_route_input() and to look at skb->dst->dev, which is
+ * changed to the destination device if ip_route_input() succeeds.
+ *
+ * Let's first consider the case that ip_route_input() succeeds:
+ *
+ * If the output device equals the logical bridge device the packet
+ * came in on, we can consider this bridging. The corresponding MAC
+ * address will be obtained in br_nf_pre_routing_finish_bridge.
+ * Otherwise, the packet is considered to be routed and we just
+ * change the destination MAC address so that the packet will
+ * later be passed up to the IP stack to be routed. For a redirected
+ * packet, ip_route_input() will give back the localhost as output device,
+ * which differs from the bridge device.
+ *
+ * Let's now consider the case that ip_route_input() fails:
+ *
+ * This can be because the destination address is martian, in which case
+ * the packet will be dropped.
+ * If IP forwarding is disabled, ip_route_input() will fail, while
+ * ip_route_output_key() can return success. The source
+ * address for ip_route_output_key() is set to zero, so ip_route_output_key()
+ * thinks we're handling a locally generated packet and won't care
+ * if IP forwarding is enabled. If the output device equals the logical bridge
+ * device, we proceed as if ip_route_input() succeeded. If it differs from the
+ * logical bridge port or if ip_route_output_key() fails we drop the packet.
+ */
+static int br_nf_pre_routing_finish(struct sock *sk, struct sk_buff *skb)
+{
+       struct net_device *dev = skb->dev;
+       struct iphdr *iph = ip_hdr(skb);
+       struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
+       struct rtable *rt;
+       int err;
+
+       nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
+
+       if (nf_bridge->pkt_otherhost) {
+               skb->pkt_type = PACKET_OTHERHOST;
+               nf_bridge->pkt_otherhost = false;
+       }
+       nf_bridge->mask &= ~BRNF_NF_BRIDGE_PREROUTING;
+       if (daddr_was_changed(skb, nf_bridge)) {
+               if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
+                       struct in_device *in_dev = __in_dev_get_rcu(dev);
+
+                       /* If err equals -EHOSTUNREACH the error is due to a
+                        * martian destination or due to the fact that
+                        * forwarding is disabled. For most martian packets,
+                        * ip_route_output_key() will fail. It won't fail for 2 types of
+                        * martian destinations: loopback destinations and destination
+                        * 0.0.0.0. In both cases the packet will be dropped because the
+                        * destination is the loopback device and not the bridge. */
+                       if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
+                               goto free_skb;
+
+                       rt = ip_route_output(dev_net(dev), iph->daddr, 0,
+                                            RT_TOS(iph->tos), 0);
+                       if (!IS_ERR(rt)) {
+                               /* - Bridged-and-DNAT'ed traffic doesn't
+                                *   require ip_forwarding. */
+                               if (rt->dst.dev == dev) {
+                                       skb_dst_set(skb, &rt->dst);
+                                       goto bridged_dnat;
+                               }
+                               ip_rt_put(rt);
+                       }
+free_skb:
+                       kfree_skb(skb);
+                       return 0;
+               } else {
+                       if (skb_dst(skb)->dev == dev) {
+bridged_dnat:
+                               skb->dev = nf_bridge->physindev;
+                               nf_bridge_update_protocol(skb);
+                               nf_bridge_push_encap_header(skb);
+                               NF_HOOK_THRESH(NFPROTO_BRIDGE,
+                                              NF_BR_PRE_ROUTING,
+                                              sk, skb, skb->dev, NULL,
+                                              br_nf_pre_routing_finish_bridge,
+                                              1);
+                               return 0;
+                       }
+                       ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr);
+                       skb->pkt_type = PACKET_HOST;
+               }
+       } else {
+               rt = bridge_parent_rtable(nf_bridge->physindev);
+               if (!rt) {
+                       kfree_skb(skb);
+                       return 0;
+               }
+               skb_dst_set_noref(skb, &rt->dst);
+       }
+
+       skb->dev = nf_bridge->physindev;
+       nf_bridge_update_protocol(skb);
+       nf_bridge_push_encap_header(skb);
+       NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, sk, skb,
+                      skb->dev, NULL,
+                      br_handle_frame_finish, 1);
+
+       return 0;
+}
+
+static struct net_device *brnf_get_logical_dev(struct sk_buff *skb, const struct net_device *dev)
+{
+       struct net_device *vlan, *br;
+
+       br = bridge_parent(dev);
+       if (brnf_pass_vlan_indev == 0 || !skb_vlan_tag_present(skb))
+               return br;
+
+       vlan = __vlan_find_dev_deep_rcu(br, skb->vlan_proto,
+                                   skb_vlan_tag_get(skb) & VLAN_VID_MASK);
+
+       return vlan ? vlan : br;
+}
+
+/* Some common code for IPv4/IPv6 */
+static struct net_device *setup_pre_routing(struct sk_buff *skb)
+{
+       struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
+
+       if (skb->pkt_type == PACKET_OTHERHOST) {
+               skb->pkt_type = PACKET_HOST;
+               nf_bridge->pkt_otherhost = true;
+       }
+
+       nf_bridge->mask |= BRNF_NF_BRIDGE_PREROUTING;
+       nf_bridge->physindev = skb->dev;
+       skb->dev = brnf_get_logical_dev(skb, skb->dev);
+
+       if (skb->protocol == htons(ETH_P_8021Q))
+               nf_bridge->orig_proto = BRNF_PROTO_8021Q;
+       else if (skb->protocol == htons(ETH_P_PPP_SES))
+               nf_bridge->orig_proto = BRNF_PROTO_PPPOE;
+
+       /* Must drop socket now because of tproxy. */
+       skb_orphan(skb);
+       return skb->dev;
+}
+
+/* Replicate the checks that IPv6 does on packet reception and pass the packet
+ * to ip6tables.
+ */
+static unsigned int br_nf_pre_routing_ipv6(const struct nf_hook_ops *ops,
+                                          struct sk_buff *skb,
+                                          const struct nf_hook_state *state)
+{
+       struct nf_bridge_info *nf_bridge;
+
+       if (br_validate_ipv6(skb))
+               return NF_DROP;
+
+       nf_bridge_put(skb->nf_bridge);
+       if (!nf_bridge_alloc(skb))
+               return NF_DROP;
+       if (!setup_pre_routing(skb))
+               return NF_DROP;
+
+       nf_bridge = nf_bridge_info_get(skb);
+       nf_bridge->ipv6_daddr = ipv6_hdr(skb)->daddr;
+
+       skb->protocol = htons(ETH_P_IPV6);
+       NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, state->sk, skb,
+               skb->dev, NULL,
+               br_nf_pre_routing_finish_ipv6);
+
+       return NF_STOLEN;
+}
+
+/* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
+ * Replicate the checks that IPv4 does on packet reception.
+ * Set skb->dev to the bridge device (i.e. parent of the
+ * receiving device) to make netfilter happy, the REDIRECT
+ * target in particular.  Save the original destination IP
+ * address to be able to detect DNAT afterwards. */
+static unsigned int br_nf_pre_routing(const struct nf_hook_ops *ops,
+                                     struct sk_buff *skb,
+                                     const struct nf_hook_state *state)
+{
+       struct nf_bridge_info *nf_bridge;
+       struct net_bridge_port *p;
+       struct net_bridge *br;
+       __u32 len = nf_bridge_encap_header_len(skb);
+
+       if (unlikely(!pskb_may_pull(skb, len)))
+               return NF_DROP;
+
+       p = br_port_get_rcu(state->in);
+       if (p == NULL)
+               return NF_DROP;
+       br = p->br;
+
+       if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb)) {
+               if (!brnf_call_ip6tables && !br->nf_call_ip6tables)
+                       return NF_ACCEPT;
+
+               nf_bridge_pull_encap_header_rcsum(skb);
+               return br_nf_pre_routing_ipv6(ops, skb, state);
+       }
+
+       if (!brnf_call_iptables && !br->nf_call_iptables)
+               return NF_ACCEPT;
+
+       if (!IS_IP(skb) && !IS_VLAN_IP(skb) && !IS_PPPOE_IP(skb))
+               return NF_ACCEPT;
+
+       nf_bridge_pull_encap_header_rcsum(skb);
+
+       if (br_validate_ipv4(skb))
+               return NF_DROP;
+
+       nf_bridge_put(skb->nf_bridge);
+       if (!nf_bridge_alloc(skb))
+               return NF_DROP;
+       if (!setup_pre_routing(skb))
+               return NF_DROP;
+
+       nf_bridge = nf_bridge_info_get(skb);
+       nf_bridge->ipv4_daddr = ip_hdr(skb)->daddr;
+
+       skb->protocol = htons(ETH_P_IP);
+
+       NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, state->sk, skb,
+               skb->dev, NULL,
+               br_nf_pre_routing_finish);
+
+       return NF_STOLEN;
+}
+
+
+/* PF_BRIDGE/LOCAL_IN ************************************************/
+/* The packet is locally destined, which requires a real
+ * dst_entry, so detach the fake one.  On the way up, the
+ * packet would pass through PRE_ROUTING again (which already
+ * took place when the packet entered the bridge), but we
+ * register an IPv4 PRE_ROUTING 'sabotage' hook that will
+ * prevent this from happening. */
+static unsigned int br_nf_local_in(const struct nf_hook_ops *ops,
+                                  struct sk_buff *skb,
+                                  const struct nf_hook_state *state)
+{
+       br_drop_fake_rtable(skb);
+       return NF_ACCEPT;
+}
+
+/* PF_BRIDGE/FORWARD *************************************************/
+static int br_nf_forward_finish(struct sock *sk, struct sk_buff *skb)
+{
+       struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
+       struct net_device *in;
+
+       if (!IS_ARP(skb) && !IS_VLAN_ARP(skb)) {
+
+               if (skb->protocol == htons(ETH_P_IP))
+                       nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
+
+               if (skb->protocol == htons(ETH_P_IPV6))
+                       nf_bridge->frag_max_size = IP6CB(skb)->frag_max_size;
+
+               in = nf_bridge->physindev;
+               if (nf_bridge->pkt_otherhost) {
+                       skb->pkt_type = PACKET_OTHERHOST;
+                       nf_bridge->pkt_otherhost = false;
+               }
+               nf_bridge_update_protocol(skb);
+       } else {
+               in = *((struct net_device **)(skb->cb));
+       }
+       nf_bridge_push_encap_header(skb);
+
+       NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_FORWARD, sk, skb,
+                      in, skb->dev, br_forward_finish, 1);
+       return 0;
+}
+
+
+/* This is the 'purely bridged' case.  For IP, we pass the packet to
+ * netfilter with indev and outdev set to the bridge device,
+ * but we are still able to filter on the 'real' indev/outdev
+ * because of the physdev module. For ARP, indev and outdev are the
+ * bridge ports. */
+static unsigned int br_nf_forward_ip(const struct nf_hook_ops *ops,
+                                    struct sk_buff *skb,
+                                    const struct nf_hook_state *state)
+{
+       struct nf_bridge_info *nf_bridge;
+       struct net_device *parent;
+       u_int8_t pf;
+
+       if (!skb->nf_bridge)
+               return NF_ACCEPT;
+
+       /* Need exclusive nf_bridge_info since we might have multiple
+        * different physoutdevs. */
+       if (!nf_bridge_unshare(skb))
+               return NF_DROP;
+
+       nf_bridge = nf_bridge_info_get(skb);
+       if (!nf_bridge)
+               return NF_DROP;
+
+       parent = bridge_parent(state->out);
+       if (!parent)
+               return NF_DROP;
+
+       if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
+               pf = NFPROTO_IPV4;
+       else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
+               pf = NFPROTO_IPV6;
+       else
+               return NF_ACCEPT;
+
+       nf_bridge_pull_encap_header(skb);
+
+       if (skb->pkt_type == PACKET_OTHERHOST) {
+               skb->pkt_type = PACKET_HOST;
+               nf_bridge->pkt_otherhost = true;
+       }
+
+       if (pf == NFPROTO_IPV4) {
+               if (br_validate_ipv4(skb))
+                       return NF_DROP;
+               IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
+       }
+
+       if (pf == NFPROTO_IPV6) {
+               if (br_validate_ipv6(skb))
+                       return NF_DROP;
+               IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
+       }
+
+       nf_bridge->physoutdev = skb->dev;
+       if (pf == NFPROTO_IPV4)
+               skb->protocol = htons(ETH_P_IP);
+       else
+               skb->protocol = htons(ETH_P_IPV6);
+
+       NF_HOOK(pf, NF_INET_FORWARD, NULL, skb,
+               brnf_get_logical_dev(skb, state->in),
+               parent, br_nf_forward_finish);
+
+       return NF_STOLEN;
+}
+
+static unsigned int br_nf_forward_arp(const struct nf_hook_ops *ops,
+                                     struct sk_buff *skb,
+                                     const struct nf_hook_state *state)
+{
+       struct net_bridge_port *p;
+       struct net_bridge *br;
+       struct net_device **d = (struct net_device **)(skb->cb);
+
+       p = br_port_get_rcu(state->out);
+       if (p == NULL)
+               return NF_ACCEPT;
+       br = p->br;
+
+       if (!brnf_call_arptables && !br->nf_call_arptables)
+               return NF_ACCEPT;
+
+       if (!IS_ARP(skb)) {
+               if (!IS_VLAN_ARP(skb))
+                       return NF_ACCEPT;
+               nf_bridge_pull_encap_header(skb);
+       }
+
+       if (arp_hdr(skb)->ar_pln != 4) {
+               if (IS_VLAN_ARP(skb))
+                       nf_bridge_push_encap_header(skb);
+               return NF_ACCEPT;
+       }
+       *d = state->in;
+       NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, state->sk, skb,
+               state->in, state->out, br_nf_forward_finish);
+
+       return NF_STOLEN;
+}
+
+#if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4) || IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
+static int br_nf_push_frag_xmit(struct sock *sk, struct sk_buff *skb)
+{
+       struct brnf_frag_data *data;
+       int err;
+
+       data = this_cpu_ptr(&brnf_frag_data_storage);
+       err = skb_cow_head(skb, data->size);
+
+       if (err) {
+               kfree_skb(skb);
+               return 0;
+       }
+
+       if (data->vlan_tci) {
+               skb->vlan_tci = data->vlan_tci;
+               skb->vlan_proto = data->vlan_proto;
+       }
+
+       skb_copy_to_linear_data_offset(skb, -data->size, data->mac, data->size);
+       __skb_push(skb, data->encap_size);
+
+       nf_bridge_info_free(skb);
+       return br_dev_queue_push_xmit(sk, skb);
+}
+#endif
+
+static int br_nf_ip_fragment(struct sock *sk, struct sk_buff *skb,
+                            int (*output)(struct sock *, struct sk_buff *))
+{
+       unsigned int mtu = ip_skb_dst_mtu(skb);
+       struct iphdr *iph = ip_hdr(skb);
+       struct rtable *rt = skb_rtable(skb);
+       struct net_device *dev = rt->dst.dev;
+
+       if (unlikely(((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) ||
+                    (IPCB(skb)->frag_max_size &&
+                     IPCB(skb)->frag_max_size > mtu))) {
+               IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
+               kfree_skb(skb);
+               return -EMSGSIZE;
+       }
+
+       return ip_do_fragment(sk, skb, output);
+}
+
+static unsigned int nf_bridge_mtu_reduction(const struct sk_buff *skb)
+{
+       if (skb->nf_bridge->orig_proto == BRNF_PROTO_PPPOE)
+               return PPPOE_SES_HLEN;
+       return 0;
+}
+
+static int br_nf_dev_queue_xmit(struct sock *sk, struct sk_buff *skb)
+{
+       struct nf_bridge_info *nf_bridge;
+       unsigned int mtu_reserved;
+
+       mtu_reserved = nf_bridge_mtu_reduction(skb);
+
+       if (skb_is_gso(skb) || skb->len + mtu_reserved <= skb->dev->mtu) {
+               nf_bridge_info_free(skb);
+               return br_dev_queue_push_xmit(sk, skb);
+       }
+
+       nf_bridge = nf_bridge_info_get(skb);
+
+#if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4)
+       /* This is wrong! We should preserve the original fragment
+        * boundaries by preserving frag_list rather than refragmenting.
+        */
+       if (skb->protocol == htons(ETH_P_IP)) {
+               struct brnf_frag_data *data;
+
+               if (br_validate_ipv4(skb))
+                       return NF_DROP;
+
+               IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
+
+               nf_bridge_update_protocol(skb);
+
+               data = this_cpu_ptr(&brnf_frag_data_storage);
+
+               data->vlan_tci = skb->vlan_tci;
+               data->vlan_proto = skb->vlan_proto;
+               data->encap_size = nf_bridge_encap_header_len(skb);
+               data->size = ETH_HLEN + data->encap_size;
+
+               skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
+                                                data->size);
+
+               return br_nf_ip_fragment(sk, skb, br_nf_push_frag_xmit);
+       }
+#endif
+#if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
+       if (skb->protocol == htons(ETH_P_IPV6)) {
+               const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops();
+               struct brnf_frag_data *data;
+
+               if (br_validate_ipv6(skb))
+                       return NF_DROP;
+
+               IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
+
+               nf_bridge_update_protocol(skb);
+
+               data = this_cpu_ptr(&brnf_frag_data_storage);
+               data->encap_size = nf_bridge_encap_header_len(skb);
+               data->size = ETH_HLEN + data->encap_size;
+
+               skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
+                                                data->size);
+
+               if (v6ops)
+                       return v6ops->fragment(sk, skb, br_nf_push_frag_xmit);
+               else
+                       return -EMSGSIZE;
+       }
+#endif
+       nf_bridge_info_free(skb);
+       return br_dev_queue_push_xmit(sk, skb);
+}
+
+/* PF_BRIDGE/POST_ROUTING ********************************************/
+static unsigned int br_nf_post_routing(const struct nf_hook_ops *ops,
+                                      struct sk_buff *skb,
+                                      const struct nf_hook_state *state)
+{
+       struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
+       struct net_device *realoutdev = bridge_parent(skb->dev);
+       u_int8_t pf;
+
+       /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
+        * on a bridge, but was delivered locally and is now being routed:
+        *
+        * POST_ROUTING was already invoked from the ip stack.
+        */
+       if (!nf_bridge || !nf_bridge->physoutdev)
+               return NF_ACCEPT;
+
+       if (!realoutdev)
+               return NF_DROP;
+
+       if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
+               pf = NFPROTO_IPV4;
+       else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
+               pf = NFPROTO_IPV6;
+       else
+               return NF_ACCEPT;
+
+       /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
+        * about the value of skb->pkt_type. */
+       if (skb->pkt_type == PACKET_OTHERHOST) {
+               skb->pkt_type = PACKET_HOST;
+               nf_bridge->pkt_otherhost = true;
+       }
+
+       nf_bridge_pull_encap_header(skb);
+       if (pf == NFPROTO_IPV4)
+               skb->protocol = htons(ETH_P_IP);
+       else
+               skb->protocol = htons(ETH_P_IPV6);
+
+       NF_HOOK(pf, NF_INET_POST_ROUTING, state->sk, skb,
+               NULL, realoutdev,
+               br_nf_dev_queue_xmit);
+
+       return NF_STOLEN;
+}
+
+/* IP/SABOTAGE *****************************************************/
+/* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
+ * for the second time. */
+static unsigned int ip_sabotage_in(const struct nf_hook_ops *ops,
+                                  struct sk_buff *skb,
+                                  const struct nf_hook_state *state)
+{
+       if (skb->nf_bridge &&
+           !(skb->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)) {
+               return NF_STOP;
+       }
+
+       return NF_ACCEPT;
+}
+
+/* This is called when br_netfilter has called into iptables/netfilter,
+ * and DNAT has taken place on a bridge-forwarded packet.
+ *
+ * neigh->output has created a new MAC header, with local br0 MAC
+ * as saddr.
+ *
+ * This restores the original MAC saddr of the bridged packet
+ * before invoking bridge forward logic to transmit the packet.
+ */
+static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff *skb)
+{
+       struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
+
+       skb_pull(skb, ETH_HLEN);
+       nf_bridge->mask &= ~BRNF_BRIDGED_DNAT;
+
+       BUILD_BUG_ON(sizeof(nf_bridge->neigh_header) != (ETH_HLEN - ETH_ALEN));
+
+       skb_copy_to_linear_data_offset(skb, -(ETH_HLEN - ETH_ALEN),
+                                      nf_bridge->neigh_header,
+                                      ETH_HLEN - ETH_ALEN);
+       skb->dev = nf_bridge->physindev;
+
+       nf_bridge->physoutdev = NULL;
+       br_handle_frame_finish(NULL, skb);
+}
+
+static int br_nf_dev_xmit(struct sk_buff *skb)
+{
+       if (skb->nf_bridge && (skb->nf_bridge->mask & BRNF_BRIDGED_DNAT)) {
+               br_nf_pre_routing_finish_bridge_slow(skb);
+               return 1;
+       }
+       return 0;
+}
+
+static const struct nf_br_ops br_ops = {
+       .br_dev_xmit_hook =     br_nf_dev_xmit,
+};
+
+void br_netfilter_enable(void)
+{
+}
+EXPORT_SYMBOL_GPL(br_netfilter_enable);
+
+/* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
+ * br_dev_queue_push_xmit is called afterwards */
+static struct nf_hook_ops br_nf_ops[] __read_mostly = {
+       {
+               .hook = br_nf_pre_routing,
+               .owner = THIS_MODULE,
+               .pf = NFPROTO_BRIDGE,
+               .hooknum = NF_BR_PRE_ROUTING,
+               .priority = NF_BR_PRI_BRNF,
+       },
+       {
+               .hook = br_nf_local_in,
+               .owner = THIS_MODULE,
+               .pf = NFPROTO_BRIDGE,
+               .hooknum = NF_BR_LOCAL_IN,
+               .priority = NF_BR_PRI_BRNF,
+       },
+       {
+               .hook = br_nf_forward_ip,
+               .owner = THIS_MODULE,
+               .pf = NFPROTO_BRIDGE,
+               .hooknum = NF_BR_FORWARD,
+               .priority = NF_BR_PRI_BRNF - 1,
+       },
+       {
+               .hook = br_nf_forward_arp,
+               .owner = THIS_MODULE,
+               .pf = NFPROTO_BRIDGE,
+               .hooknum = NF_BR_FORWARD,
+               .priority = NF_BR_PRI_BRNF,
+       },
+       {
+               .hook = br_nf_post_routing,
+               .owner = THIS_MODULE,
+               .pf = NFPROTO_BRIDGE,
+               .hooknum = NF_BR_POST_ROUTING,
+               .priority = NF_BR_PRI_LAST,
+       },
+       {
+               .hook = ip_sabotage_in,
+               .owner = THIS_MODULE,
+               .pf = NFPROTO_IPV4,
+               .hooknum = NF_INET_PRE_ROUTING,
+               .priority = NF_IP_PRI_FIRST,
+       },
+       {
+               .hook = ip_sabotage_in,
+               .owner = THIS_MODULE,
+               .pf = NFPROTO_IPV6,
+               .hooknum = NF_INET_PRE_ROUTING,
+               .priority = NF_IP6_PRI_FIRST,
+       },
+};
+
+#ifdef CONFIG_SYSCTL
+static
+int brnf_sysctl_call_tables(struct ctl_table *ctl, int write,
+                           void __user *buffer, size_t *lenp, loff_t *ppos)
+{
+       int ret;
+
+       ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
+
+       if (write && *(int *)(ctl->data))
+               *(int *)(ctl->data) = 1;
+       return ret;
+}
+
+static struct ctl_table brnf_table[] = {
+       {
+               .procname       = "bridge-nf-call-arptables",
+               .data           = &brnf_call_arptables,
+               .maxlen         = sizeof(int),
+               .mode           = 0644,
+               .proc_handler   = brnf_sysctl_call_tables,
+       },
+       {
+               .procname       = "bridge-nf-call-iptables",
+               .data           = &brnf_call_iptables,
+               .maxlen         = sizeof(int),
+               .mode           = 0644,
+               .proc_handler   = brnf_sysctl_call_tables,
+       },
+       {
+               .procname       = "bridge-nf-call-ip6tables",
+               .data           = &brnf_call_ip6tables,
+               .maxlen         = sizeof(int),
+               .mode           = 0644,
+               .proc_handler   = brnf_sysctl_call_tables,
+       },
+       {
+               .procname       = "bridge-nf-filter-vlan-tagged",
+               .data           = &brnf_filter_vlan_tagged,
+               .maxlen         = sizeof(int),
+               .mode           = 0644,
+               .proc_handler   = brnf_sysctl_call_tables,
+       },
+       {
+               .procname       = "bridge-nf-filter-pppoe-tagged",
+               .data           = &brnf_filter_pppoe_tagged,
+               .maxlen         = sizeof(int),
+               .mode           = 0644,
+               .proc_handler   = brnf_sysctl_call_tables,
+       },
+       {
+               .procname       = "bridge-nf-pass-vlan-input-dev",
+               .data           = &brnf_pass_vlan_indev,
+               .maxlen         = sizeof(int),
+               .mode           = 0644,
+               .proc_handler   = brnf_sysctl_call_tables,
+       },
+       { }
+};
+#endif
+
+static int __init br_netfilter_init(void)
+{
+       int ret;
+
+       ret = nf_register_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
+       if (ret < 0)
+               return ret;
+
+#ifdef CONFIG_SYSCTL
+       brnf_sysctl_header = register_net_sysctl(&init_net, "net/bridge", brnf_table);
+       if (brnf_sysctl_header == NULL) {
+               printk(KERN_WARNING
+                      "br_netfilter: can't register to sysctl.\n");
+               nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
+               return -ENOMEM;
+       }
+#endif
+       RCU_INIT_POINTER(nf_br_ops, &br_ops);
+       printk(KERN_NOTICE "Bridge firewalling registered\n");
+       return 0;
+}
+
+static void __exit br_netfilter_fini(void)
+{
+       RCU_INIT_POINTER(nf_br_ops, NULL);
+       nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
+#ifdef CONFIG_SYSCTL
+       unregister_net_sysctl_table(brnf_sysctl_header);
+#endif
+}
+
+module_init(br_netfilter_init);
+module_exit(br_netfilter_fini);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
+MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
+MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");