net/bridge/br_netfilter_hooks.c

   1 /*
   2  *      Handle firewalling
   3  *      Linux ethernet bridge
   4  *
   5  *      Authors:
   6  *      Lennert Buytenhek               <buytenh@gnu.org>
   7  *      Bart De Schuymer                <bdschuym@pandora.be>
   8  *
   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.
  13  *
  14  *      Lennert dedicates this file to Kerstin Wurdinger.
  15  */
  16
  17 #include <linux/module.h>
  18 #include <linux/kernel.h>
  19 #include <linux/slab.h>
  20 #include <linux/ip.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>
  34
  35 #include <net/ip.h>
  36 #include <net/ipv6.h>
  37 #include <net/addrconf.h>
  38 #include <net/route.h>
  39 #include <net/netfilter/br_netfilter.h>
  40
  41 #include <asm/uaccess.h>
  42 #include "br_private.h"
  43 #ifdef CONFIG_SYSCTL
  44 #include <linux/sysctl.h>
  45 #endif
  46
  47 #ifdef CONFIG_SYSCTL
  48 static struct ctl_table_header *brnf_sysctl_header;
  49 static int brnf_call_iptables __read_mostly = 1;
  50 static int brnf_call_ip6tables __read_mostly = 1;
  51 static int brnf_call_arptables __read_mostly = 1;
  52 static int brnf_filter_vlan_tagged __read_mostly = 0;
  53 static int brnf_filter_pppoe_tagged __read_mostly = 0;
  54 static int brnf_pass_vlan_indev __read_mostly = 0;
  55 #else
  56 #define brnf_call_iptables 1
  57 #define brnf_call_ip6tables 1
  58 #define brnf_call_arptables 1
  59 #define brnf_filter_vlan_tagged 0
  60 #define brnf_filter_pppoe_tagged 0
  61 #define brnf_pass_vlan_indev 0
  62 #endif
  63
  64 #define IS_IP(skb) \
  65         (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP))
  66
  67 #define IS_IPV6(skb) \
  68         (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6))
  69
  70 #define IS_ARP(skb) \
  71         (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP))
  72
  73 static inline __be16 vlan_proto(const struct sk_buff *skb)
  74 {
  75         if (skb_vlan_tag_present(skb))
  76                 return skb->protocol;
  77         else if (skb->protocol == htons(ETH_P_8021Q))
  78                 return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
  79         else
  80                 return 0;
  81 }
  82
  83 #define IS_VLAN_IP(skb) \
  84         (vlan_proto(skb) == htons(ETH_P_IP) && \
  85          brnf_filter_vlan_tagged)
  86
  87 #define IS_VLAN_IPV6(skb) \
  88         (vlan_proto(skb) == htons(ETH_P_IPV6) && \
  89          brnf_filter_vlan_tagged)
  90
  91 #define IS_VLAN_ARP(skb) \
  92         (vlan_proto(skb) == htons(ETH_P_ARP) && \
  93          brnf_filter_vlan_tagged)
  94
  95 static inline __be16 pppoe_proto(const struct sk_buff *skb)
  96 {
  97         return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
  98                             sizeof(struct pppoe_hdr)));
  99 }
 100
 101 #define IS_PPPOE_IP(skb) \
 102         (skb->protocol == htons(ETH_P_PPP_SES) && \
 103          pppoe_proto(skb) == htons(PPP_IP) && \
 104          brnf_filter_pppoe_tagged)
 105
 106 #define IS_PPPOE_IPV6(skb) \
 107         (skb->protocol == htons(ETH_P_PPP_SES) && \
 108          pppoe_proto(skb) == htons(PPP_IPV6) && \
 109          brnf_filter_pppoe_tagged)
 110
 111 /* largest possible L2 header, see br_nf_dev_queue_xmit() */
 112 #define NF_BRIDGE_MAX_MAC_HEADER_LENGTH (PPPOE_SES_HLEN + ETH_HLEN)
 113
 114 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4)
 115 struct brnf_frag_data {
 116         char mac[NF_BRIDGE_MAX_MAC_HEADER_LENGTH];
 117         u8 encap_size;
 118         u8 size;
 119         u16 vlan_tci;
 120         __be16 vlan_proto;
 121 };
 122
 123 static DEFINE_PER_CPU(struct brnf_frag_data, brnf_frag_data_storage);
 124 #endif
 125
 126 static struct nf_bridge_info *nf_bridge_info_get(const struct sk_buff *skb)
 127 {
 128         return skb->nf_bridge;
 129 }
 130
 131 static void nf_bridge_info_free(struct sk_buff *skb)
 132 {
 133         if (skb->nf_bridge) {
 134                 nf_bridge_put(skb->nf_bridge);
 135                 skb->nf_bridge = NULL;
 136         }
 137 }
 138
 139 static inline struct rtable *bridge_parent_rtable(const struct net_device *dev)
 140 {
 141         struct net_bridge_port *port;
 142
 143         port = br_port_get_rcu(dev);
 144         return port ? &port->br->fake_rtable : NULL;
 145 }
 146
 147 static inline struct net_device *bridge_parent(const struct net_device *dev)
 148 {
 149         struct net_bridge_port *port;
 150
 151         port = br_port_get_rcu(dev);
 152         return port ? port->br->dev : NULL;
 153 }
 154
 155 static inline struct nf_bridge_info *nf_bridge_alloc(struct sk_buff *skb)
 156 {
 157         skb->nf_bridge = kzalloc(sizeof(struct nf_bridge_info), GFP_ATOMIC);
 158         if (likely(skb->nf_bridge))
 159                 atomic_set(&(skb->nf_bridge->use), 1);
 160
 161         return skb->nf_bridge;
 162 }
 163
 164 static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
 165 {
 166         struct nf_bridge_info *nf_bridge = skb->nf_bridge;
 167
 168         if (atomic_read(&nf_bridge->use) > 1) {
 169                 struct nf_bridge_info *tmp = nf_bridge_alloc(skb);
 170
 171                 if (tmp) {
 172                         memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info));
 173                         atomic_set(&tmp->use, 1);
 174                 }
 175                 nf_bridge_put(nf_bridge);
 176                 nf_bridge = tmp;
 177         }
 178         return nf_bridge;
 179 }
 180
 181 static unsigned int nf_bridge_encap_header_len(const struct sk_buff *skb)
 182 {
 183         switch (skb->protocol) {
 184         case __cpu_to_be16(ETH_P_8021Q):
 185                 return VLAN_HLEN;
 186         case __cpu_to_be16(ETH_P_PPP_SES):
 187                 return PPPOE_SES_HLEN;
 188         default:
 189                 return 0;
 190         }
 191 }
 192
 193 static inline void nf_bridge_push_encap_header(struct sk_buff *skb)
 194 {
 195         unsigned int len = nf_bridge_encap_header_len(skb);
 196
 197         skb_push(skb, len);
 198         skb->network_header -= len;
 199 }
 200
 201 static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
 202 {
 203         unsigned int len = nf_bridge_encap_header_len(skb);
 204
 205         skb_pull(skb, len);
 206         skb->network_header += len;
 207 }
 208
 209 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
 210 {
 211         unsigned int len = nf_bridge_encap_header_len(skb);
 212
 213         skb_pull_rcsum(skb, len);
 214         skb->network_header += len;
 215 }
 216
 217 /* When handing a packet over to the IP layer
 218  * check whether we have a skb that is in the
 219  * expected format
 220  */
 221
 222 static int br_validate_ipv4(struct sk_buff *skb)
 223 {
 224         const struct iphdr *iph;
 225         struct net_device *dev = skb->dev;
 226         u32 len;
 227
 228         if (!pskb_may_pull(skb, sizeof(struct iphdr)))
 229                 goto inhdr_error;
 230
 231         iph = ip_hdr(skb);
 232
 233         /* Basic sanity checks */
 234         if (iph->ihl < 5 || iph->version != 4)
 235                 goto inhdr_error;
 236
 237         if (!pskb_may_pull(skb, iph->ihl*4))
 238                 goto inhdr_error;
 239
 240         iph = ip_hdr(skb);
 241         if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
 242                 goto inhdr_error;
 243
 244         len = ntohs(iph->tot_len);
 245         if (skb->len < len) {
 246                 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INTRUNCATEDPKTS);
 247                 goto drop;
 248         } else if (len < (iph->ihl*4))
 249                 goto inhdr_error;
 250
 251         if (pskb_trim_rcsum(skb, len)) {
 252                 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
 253                 goto drop;
 254         }
 255
 256         memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
 257         /* We should really parse IP options here but until
 258          * somebody who actually uses IP options complains to
 259          * us we'll just silently ignore the options because
 260          * we're lazy!
 261          */
 262         return 0;
 263
 264 inhdr_error:
 265         IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
 266 drop:
 267         return -1;
 268 }
 269
 270 /* We only check the length. A bridge shouldn't do any hop-by-hop stuff
 271  * anyway
 272  */
 273 static int check_hbh_len(struct sk_buff *skb)
 274 {
 275         unsigned char *raw = (u8 *)(ipv6_hdr(skb) + 1);
 276         u32 pkt_len;
 277         const unsigned char *nh = skb_network_header(skb);
 278         int off = raw - nh;
 279         int len = (raw[1] + 1) << 3;
 280
 281         if ((raw + len) - skb->data > skb_headlen(skb))
 282                 goto bad;
 283
 284         off += 2;
 285         len -= 2;
 286
 287         while (len > 0) {
 288                 int optlen = nh[off + 1] + 2;
 289
 290                 switch (nh[off]) {
 291                 case IPV6_TLV_PAD1:
 292                         optlen = 1;
 293                         break;
 294
 295                 case IPV6_TLV_PADN:
 296                         break;
 297
 298                 case IPV6_TLV_JUMBO:
 299                         if (nh[off + 1] != 4 || (off & 3) != 2)
 300                                 goto bad;
 301                         pkt_len = ntohl(*(__be32 *)(nh + off + 2));
 302                         if (pkt_len <= IPV6_MAXPLEN ||
 303                             ipv6_hdr(skb)->payload_len)
 304                                 goto bad;
 305                         if (pkt_len > skb->len - sizeof(struct ipv6hdr))
 306                                 goto bad;
 307                         if (pskb_trim_rcsum(skb,
 308                                             pkt_len + sizeof(struct ipv6hdr)))
 309                                 goto bad;
 310                         nh = skb_network_header(skb);
 311                         break;
 312                 default:
 313                         if (optlen > len)
 314                                 goto bad;
 315                         break;
 316                 }
 317                 off += optlen;
 318                 len -= optlen;
 319         }
 320         if (len == 0)
 321                 return 0;
 322 bad:
 323         return -1;
 324 }
 325
 326 /* Equivalent to br_validate_ipv4 for IPv6 */
 327 static int br_validate_ipv6(struct sk_buff *skb)
 328 {
 329         const struct ipv6hdr *hdr;
 330         struct net_device *dev = skb->dev;
 331         struct inet6_dev *idev = in6_dev_get(skb->dev);
 332         u32 pkt_len;
 333         u8 ip6h_len = sizeof(struct ipv6hdr);
 334
 335         if (!pskb_may_pull(skb, ip6h_len))
 336                 goto inhdr_error;
 337
 338         if (skb->len < ip6h_len)
 339                 goto drop;
 340
 341         hdr = ipv6_hdr(skb);
 342
 343         if (hdr->version != 6)
 344                 goto inhdr_error;
 345
 346         pkt_len = ntohs(hdr->payload_len);
 347
 348         if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
 349                 if (pkt_len + ip6h_len > skb->len) {
 350                         IP6_INC_STATS_BH(dev_net(dev), idev,
 351                                          IPSTATS_MIB_INTRUNCATEDPKTS);
 352                         goto drop;
 353                 }
 354                 if (pskb_trim_rcsum(skb, pkt_len + ip6h_len)) {
 355                         IP6_INC_STATS_BH(dev_net(dev), idev,
 356                                          IPSTATS_MIB_INDISCARDS);
 357                         goto drop;
 358                 }
 359         }
 360         if (hdr->nexthdr == NEXTHDR_HOP && check_hbh_len(skb))
 361                 goto drop;
 362
 363         memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
 364         /* No IP options in IPv6 header; however it should be
 365          * checked if some next headers need special treatment
 366          */
 367         return 0;
 368
 369 inhdr_error:
 370         IP6_INC_STATS_BH(dev_net(dev), idev, IPSTATS_MIB_INHDRERRORS);
 371 drop:
 372         return -1;
 373 }
 374
 375 static void nf_bridge_update_protocol(struct sk_buff *skb)
 376 {
 377         switch (skb->nf_bridge->orig_proto) {
 378         case BRNF_PROTO_8021Q:
 379                 skb->protocol = htons(ETH_P_8021Q);
 380                 break;
 381         case BRNF_PROTO_PPPOE:
 382                 skb->protocol = htons(ETH_P_PPP_SES);
 383                 break;
 384         case BRNF_PROTO_UNCHANGED:
 385                 break;
 386         }
 387 }
 388
 389 /* Obtain the correct destination MAC address, while preserving the original
 390  * source MAC address. If we already know this address, we just copy it. If we
 391  * don't, we use the neighbour framework to find out. In both cases, we make
 392  * sure that br_handle_frame_finish() is called afterwards.
 393  */
 394 static int br_nf_pre_routing_finish_bridge(struct sock *sk, struct sk_buff *skb)
 395 {
 396         struct neighbour *neigh;
 397         struct dst_entry *dst;
 398
 399         skb->dev = bridge_parent(skb->dev);
 400         if (!skb->dev)
 401                 goto free_skb;
 402         dst = skb_dst(skb);
 403         neigh = dst_neigh_lookup_skb(dst, skb);
 404         if (neigh) {
 405                 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
 406                 int ret;
 407
 408                 if (neigh->hh.hh_len) {
 409                         neigh_hh_bridge(&neigh->hh, skb);
 410                         skb->dev = nf_bridge->physindev;
 411                         ret = br_handle_frame_finish(sk, skb);
 412                 } else {
 413                         /* the neighbour function below overwrites the complete
 414                          * MAC header, so we save the Ethernet source address and
 415                          * protocol number.
 416                          */
 417                         skb_copy_from_linear_data_offset(skb,
 418                                                          -(ETH_HLEN-ETH_ALEN),
 419                                                          nf_bridge->neigh_header,
 420                                                          ETH_HLEN-ETH_ALEN);
 421                         /* tell br_dev_xmit to continue with forwarding */
 422                         nf_bridge->mask |= BRNF_BRIDGED_DNAT;
 423                         /* FIXME Need to refragment */
 424                         ret = neigh->output(neigh, skb);
 425                 }
 426                 neigh_release(neigh);
 427                 return ret;
 428         }
 429 free_skb:
 430         kfree_skb(skb);
 431         return 0;
 432 }
 433
 434 static bool daddr_was_changed(const struct sk_buff *skb,
 435                               const struct nf_bridge_info *nf_bridge)
 436 {
 437         switch (skb->protocol) {
 438         case htons(ETH_P_IP):
 439                 return ip_hdr(skb)->daddr != nf_bridge->ipv4_daddr;
 440         case htons(ETH_P_IPV6):
 441                 return memcmp(&nf_bridge->ipv6_daddr, &ipv6_hdr(skb)->daddr,
 442                               sizeof(ipv6_hdr(skb)->daddr)) != 0;
 443         default:
 444                 return false;
 445         }
 446 }
 447
 448 /* PF_BRIDGE/PRE_ROUTING: Undo the changes made for ip6tables
 449  * PREROUTING and continue the bridge PRE_ROUTING hook. See comment
 450  * for br_nf_pre_routing_finish(), same logic is used here but
 451  * equivalent IPv6 function ip6_route_input() called indirectly.
 452  */
 453 static int br_nf_pre_routing_finish_ipv6(struct sock *sk, struct sk_buff *skb)
 454 {
 455         struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
 456         struct rtable *rt;
 457         struct net_device *dev = skb->dev;
 458         const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops();
 459
 460         nf_bridge->frag_max_size = IP6CB(skb)->frag_max_size;
 461
 462         if (nf_bridge->pkt_otherhost) {
 463                 skb->pkt_type = PACKET_OTHERHOST;
 464                 nf_bridge->pkt_otherhost = false;
 465         }
 466         nf_bridge->mask &= ~BRNF_NF_BRIDGE_PREROUTING;
 467         if (daddr_was_changed(skb, nf_bridge)) {
 468                 skb_dst_drop(skb);
 469                 v6ops->route_input(skb);
 470
 471                 if (skb_dst(skb)->error) {
 472                         kfree_skb(skb);
 473                         return 0;
 474                 }
 475
 476                 if (skb_dst(skb)->dev == dev) {
 477                         skb->dev = nf_bridge->physindev;
 478                         nf_bridge_update_protocol(skb);
 479                         nf_bridge_push_encap_header(skb);
 480                         NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING,
 481                                        sk, skb, skb->dev, NULL,
 482                                        br_nf_pre_routing_finish_bridge,
 483                                        1);
 484                         return 0;
 485                 }
 486                 ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr);
 487                 skb->pkt_type = PACKET_HOST;
 488         } else {
 489                 rt = bridge_parent_rtable(nf_bridge->physindev);
 490                 if (!rt) {
 491                         kfree_skb(skb);
 492                         return 0;
 493                 }
 494                 skb_dst_set_noref(skb, &rt->dst);
 495         }
 496
 497         skb->dev = nf_bridge->physindev;
 498         nf_bridge_update_protocol(skb);
 499         nf_bridge_push_encap_header(skb);
 500         NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, sk, skb,
 501                        skb->dev, NULL,
 502                        br_handle_frame_finish, 1);
 503
 504         return 0;
 505 }
 506
 507 /* This requires some explaining. If DNAT has taken place,
 508  * we will need to fix up the destination Ethernet address.
 509  * This is also true when SNAT takes place (for the reply direction).
 510  *
 511  * There are two cases to consider:
 512  * 1. The packet was DNAT'ed to a device in the same bridge
 513  *    port group as it was received on. We can still bridge
 514  *    the packet.
 515  * 2. The packet was DNAT'ed to a different device, either
 516  *    a non-bridged device or another bridge port group.
 517  *    The packet will need to be routed.
 518  *
 519  * The correct way of distinguishing between these two cases is to
 520  * call ip_route_input() and to look at skb->dst->dev, which is
 521  * changed to the destination device if ip_route_input() succeeds.
 522  *
 523  * Let's first consider the case that ip_route_input() succeeds:
 524  *
 525  * If the output device equals the logical bridge device the packet
 526  * came in on, we can consider this bridging. The corresponding MAC
 527  * address will be obtained in br_nf_pre_routing_finish_bridge.
 528  * Otherwise, the packet is considered to be routed and we just
 529  * change the destination MAC address so that the packet will
 530  * later be passed up to the IP stack to be routed. For a redirected
 531  * packet, ip_route_input() will give back the localhost as output device,
 532  * which differs from the bridge device.
 533  *
 534  * Let's now consider the case that ip_route_input() fails:
 535  *
 536  * This can be because the destination address is martian, in which case
 537  * the packet will be dropped.
 538  * If IP forwarding is disabled, ip_route_input() will fail, while
 539  * ip_route_output_key() can return success. The source
 540  * address for ip_route_output_key() is set to zero, so ip_route_output_key()
 541  * thinks we're handling a locally generated packet and won't care
 542  * if IP forwarding is enabled. If the output device equals the logical bridge
 543  * device, we proceed as if ip_route_input() succeeded. If it differs from the
 544  * logical bridge port or if ip_route_output_key() fails we drop the packet.
 545  */
 546 static int br_nf_pre_routing_finish(struct sock *sk, struct sk_buff *skb)
 547 {
 548         struct net_device *dev = skb->dev;
 549         struct iphdr *iph = ip_hdr(skb);
 550         struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
 551         struct rtable *rt;
 552         int err;
 553
 554         nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
 555
 556         if (nf_bridge->pkt_otherhost) {
 557                 skb->pkt_type = PACKET_OTHERHOST;
 558                 nf_bridge->pkt_otherhost = false;
 559         }
 560         nf_bridge->mask &= ~BRNF_NF_BRIDGE_PREROUTING;
 561         if (daddr_was_changed(skb, nf_bridge)) {
 562                 if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
 563                         struct in_device *in_dev = __in_dev_get_rcu(dev);
 564
 565                         /* If err equals -EHOSTUNREACH the error is due to a
 566                          * martian destination or due to the fact that
 567                          * forwarding is disabled. For most martian packets,
 568                          * ip_route_output_key() will fail. It won't fail for 2 types of
 569                          * martian destinations: loopback destinations and destination
 570                          * 0.0.0.0. In both cases the packet will be dropped because the
 571                          * destination is the loopback device and not the bridge. */
 572                         if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
 573                                 goto free_skb;
 574
 575                         rt = ip_route_output(dev_net(dev), iph->daddr, 0,
 576                                              RT_TOS(iph->tos), 0);
 577                         if (!IS_ERR(rt)) {
 578                                 /* - Bridged-and-DNAT'ed traffic doesn't
 579                                  *   require ip_forwarding. */
 580                                 if (rt->dst.dev == dev) {
 581                                         skb_dst_set(skb, &rt->dst);
 582                                         goto bridged_dnat;
 583                                 }
 584                                 ip_rt_put(rt);
 585                         }
 586 free_skb:
 587                         kfree_skb(skb);
 588                         return 0;
 589                 } else {
 590                         if (skb_dst(skb)->dev == dev) {
 591 bridged_dnat:
 592                                 skb->dev = nf_bridge->physindev;
 593                                 nf_bridge_update_protocol(skb);
 594                                 nf_bridge_push_encap_header(skb);
 595                                 NF_HOOK_THRESH(NFPROTO_BRIDGE,
 596                                                NF_BR_PRE_ROUTING,
 597                                                sk, skb, skb->dev, NULL,
 598                                                br_nf_pre_routing_finish_bridge,
 599                                                1);
 600                                 return 0;
 601                         }
 602                         ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr);
 603                         skb->pkt_type = PACKET_HOST;
 604                 }
 605         } else {
 606                 rt = bridge_parent_rtable(nf_bridge->physindev);
 607                 if (!rt) {
 608                         kfree_skb(skb);
 609                         return 0;
 610                 }
 611                 skb_dst_set_noref(skb, &rt->dst);
 612         }
 613
 614         skb->dev = nf_bridge->physindev;
 615         nf_bridge_update_protocol(skb);
 616         nf_bridge_push_encap_header(skb);
 617         NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, sk, skb,
 618                        skb->dev, NULL,
 619                        br_handle_frame_finish, 1);
 620
 621         return 0;
 622 }
 623
 624 static struct net_device *brnf_get_logical_dev(struct sk_buff *skb, const struct net_device *dev)
 625 {
 626         struct net_device *vlan, *br;
 627
 628         br = bridge_parent(dev);
 629         if (brnf_pass_vlan_indev == 0 || !skb_vlan_tag_present(skb))
 630                 return br;
 631
 632         vlan = __vlan_find_dev_deep_rcu(br, skb->vlan_proto,
 633                                     skb_vlan_tag_get(skb) & VLAN_VID_MASK);
 634
 635         return vlan ? vlan : br;
 636 }
 637
 638 /* Some common code for IPv4/IPv6 */
 639 static struct net_device *setup_pre_routing(struct sk_buff *skb)
 640 {
 641         struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
 642
 643         if (skb->pkt_type == PACKET_OTHERHOST) {
 644                 skb->pkt_type = PACKET_HOST;
 645                 nf_bridge->pkt_otherhost = true;
 646         }
 647
 648         nf_bridge->mask |= BRNF_NF_BRIDGE_PREROUTING;
 649         nf_bridge->physindev = skb->dev;
 650         skb->dev = brnf_get_logical_dev(skb, skb->dev);
 651
 652         if (skb->protocol == htons(ETH_P_8021Q))
 653                 nf_bridge->orig_proto = BRNF_PROTO_8021Q;
 654         else if (skb->protocol == htons(ETH_P_PPP_SES))
 655                 nf_bridge->orig_proto = BRNF_PROTO_PPPOE;
 656
 657         /* Must drop socket now because of tproxy. */
 658         skb_orphan(skb);
 659         return skb->dev;
 660 }
 661
 662 /* Replicate the checks that IPv6 does on packet reception and pass the packet
 663  * to ip6tables.
 664  */
 665 static unsigned int br_nf_pre_routing_ipv6(const struct nf_hook_ops *ops,
 666                                            struct sk_buff *skb,
 667                                            const struct nf_hook_state *state)
 668 {
 669         struct nf_bridge_info *nf_bridge;
 670
 671         if (br_validate_ipv6(skb))
 672                 return NF_DROP;
 673
 674         nf_bridge_put(skb->nf_bridge);
 675         if (!nf_bridge_alloc(skb))
 676                 return NF_DROP;
 677         if (!setup_pre_routing(skb))
 678                 return NF_DROP;
 679
 680         nf_bridge = nf_bridge_info_get(skb);
 681         nf_bridge->ipv6_daddr = ipv6_hdr(skb)->daddr;
 682
 683         skb->protocol = htons(ETH_P_IPV6);
 684         NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, state->sk, skb,
 685                 skb->dev, NULL,
 686                 br_nf_pre_routing_finish_ipv6);
 687
 688         return NF_STOLEN;
 689 }
 690
 691 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
 692  * Replicate the checks that IPv4 does on packet reception.
 693  * Set skb->dev to the bridge device (i.e. parent of the
 694  * receiving device) to make netfilter happy, the REDIRECT
 695  * target in particular.  Save the original destination IP
 696  * address to be able to detect DNAT afterwards. */
 697 static unsigned int br_nf_pre_routing(const struct nf_hook_ops *ops,
 698                                       struct sk_buff *skb,
 699                                       const struct nf_hook_state *state)
 700 {
 701         struct nf_bridge_info *nf_bridge;
 702         struct net_bridge_port *p;
 703         struct net_bridge *br;
 704         __u32 len = nf_bridge_encap_header_len(skb);
 705
 706         if (unlikely(!pskb_may_pull(skb, len)))
 707                 return NF_DROP;
 708
 709         p = br_port_get_rcu(state->in);
 710         if (p == NULL)
 711                 return NF_DROP;
 712         br = p->br;
 713
 714         if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb)) {
 715                 if (!brnf_call_ip6tables && !br->nf_call_ip6tables)
 716                         return NF_ACCEPT;
 717
 718                 nf_bridge_pull_encap_header_rcsum(skb);
 719                 return br_nf_pre_routing_ipv6(ops, skb, state);
 720         }
 721
 722         if (!brnf_call_iptables && !br->nf_call_iptables)
 723                 return NF_ACCEPT;
 724
 725         if (!IS_IP(skb) && !IS_VLAN_IP(skb) && !IS_PPPOE_IP(skb))
 726                 return NF_ACCEPT;
 727
 728         nf_bridge_pull_encap_header_rcsum(skb);
 729
 730         if (br_validate_ipv4(skb))
 731                 return NF_DROP;
 732
 733         nf_bridge_put(skb->nf_bridge);
 734         if (!nf_bridge_alloc(skb))
 735                 return NF_DROP;
 736         if (!setup_pre_routing(skb))
 737                 return NF_DROP;
 738
 739         nf_bridge = nf_bridge_info_get(skb);
 740         nf_bridge->ipv4_daddr = ip_hdr(skb)->daddr;
 741
 742         skb->protocol = htons(ETH_P_IP);
 743
 744         NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, state->sk, skb,
 745                 skb->dev, NULL,
 746                 br_nf_pre_routing_finish);
 747
 748         return NF_STOLEN;
 749 }
 750
 751
 752 /* PF_BRIDGE/LOCAL_IN ************************************************/
 753 /* The packet is locally destined, which requires a real
 754  * dst_entry, so detach the fake one.  On the way up, the
 755  * packet would pass through PRE_ROUTING again (which already
 756  * took place when the packet entered the bridge), but we
 757  * register an IPv4 PRE_ROUTING 'sabotage' hook that will
 758  * prevent this from happening. */
 759 static unsigned int br_nf_local_in(const struct nf_hook_ops *ops,
 760                                    struct sk_buff *skb,
 761                                    const struct nf_hook_state *state)
 762 {
 763         br_drop_fake_rtable(skb);
 764         return NF_ACCEPT;
 765 }
 766
 767 /* PF_BRIDGE/FORWARD *************************************************/
 768 static int br_nf_forward_finish(struct sock *sk, struct sk_buff *skb)
 769 {
 770         struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
 771         struct net_device *in;
 772
 773         if (!IS_ARP(skb) && !IS_VLAN_ARP(skb)) {
 774
 775                 if (skb->protocol == htons(ETH_P_IP))
 776                         nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
 777
 778                 if (skb->protocol == htons(ETH_P_IPV6))
 779                         nf_bridge->frag_max_size = IP6CB(skb)->frag_max_size;
 780
 781                 in = nf_bridge->physindev;
 782                 if (nf_bridge->pkt_otherhost) {
 783                         skb->pkt_type = PACKET_OTHERHOST;
 784                         nf_bridge->pkt_otherhost = false;
 785                 }
 786                 nf_bridge_update_protocol(skb);
 787         } else {
 788                 in = *((struct net_device **)(skb->cb));
 789         }
 790         nf_bridge_push_encap_header(skb);
 791
 792         NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_FORWARD, sk, skb,
 793                        in, skb->dev, br_forward_finish, 1);
 794         return 0;
 795 }
 796
 797
 798 /* This is the 'purely bridged' case.  For IP, we pass the packet to
 799  * netfilter with indev and outdev set to the bridge device,
 800  * but we are still able to filter on the 'real' indev/outdev
 801  * because of the physdev module. For ARP, indev and outdev are the
 802  * bridge ports. */
 803 static unsigned int br_nf_forward_ip(const struct nf_hook_ops *ops,
 804                                      struct sk_buff *skb,
 805                                      const struct nf_hook_state *state)
 806 {
 807         struct nf_bridge_info *nf_bridge;
 808         struct net_device *parent;
 809         u_int8_t pf;
 810
 811         if (!skb->nf_bridge)
 812                 return NF_ACCEPT;
 813
 814         /* Need exclusive nf_bridge_info since we might have multiple
 815          * different physoutdevs. */
 816         if (!nf_bridge_unshare(skb))
 817                 return NF_DROP;
 818
 819         nf_bridge = nf_bridge_info_get(skb);
 820         if (!nf_bridge)
 821                 return NF_DROP;
 822
 823         parent = bridge_parent(state->out);
 824         if (!parent)
 825                 return NF_DROP;
 826
 827         if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
 828                 pf = NFPROTO_IPV4;
 829         else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
 830                 pf = NFPROTO_IPV6;
 831         else
 832                 return NF_ACCEPT;
 833
 834         nf_bridge_pull_encap_header(skb);
 835
 836         if (skb->pkt_type == PACKET_OTHERHOST) {
 837                 skb->pkt_type = PACKET_HOST;
 838                 nf_bridge->pkt_otherhost = true;
 839         }
 840
 841         if (pf == NFPROTO_IPV4) {
 842                 if (br_validate_ipv4(skb))
 843                         return NF_DROP;
 844                 IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
 845         }
 846
 847         if (pf == NFPROTO_IPV6) {
 848                 if (br_validate_ipv6(skb))
 849                         return NF_DROP;
 850                 IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
 851         }
 852
 853         nf_bridge->physoutdev = skb->dev;
 854         if (pf == NFPROTO_IPV4)
 855                 skb->protocol = htons(ETH_P_IP);
 856         else
 857                 skb->protocol = htons(ETH_P_IPV6);
 858
 859         NF_HOOK(pf, NF_INET_FORWARD, NULL, skb,
 860                 brnf_get_logical_dev(skb, state->in),
 861                 parent, br_nf_forward_finish);
 862
 863         return NF_STOLEN;
 864 }
 865
 866 static unsigned int br_nf_forward_arp(const struct nf_hook_ops *ops,
 867                                       struct sk_buff *skb,
 868                                       const struct nf_hook_state *state)
 869 {
 870         struct net_bridge_port *p;
 871         struct net_bridge *br;
 872         struct net_device **d = (struct net_device **)(skb->cb);
 873
 874         p = br_port_get_rcu(state->out);
 875         if (p == NULL)
 876                 return NF_ACCEPT;
 877         br = p->br;
 878
 879         if (!brnf_call_arptables && !br->nf_call_arptables)
 880                 return NF_ACCEPT;
 881
 882         if (!IS_ARP(skb)) {
 883                 if (!IS_VLAN_ARP(skb))
 884                         return NF_ACCEPT;
 885                 nf_bridge_pull_encap_header(skb);
 886         }
 887
 888         if (arp_hdr(skb)->ar_pln != 4) {
 889                 if (IS_VLAN_ARP(skb))
 890                         nf_bridge_push_encap_header(skb);
 891                 return NF_ACCEPT;
 892         }
 893         *d = state->in;
 894         NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, state->sk, skb,
 895                 state->in, state->out, br_nf_forward_finish);
 896
 897         return NF_STOLEN;
 898 }
 899
 900 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4) || IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
 901 static int br_nf_push_frag_xmit(struct sock *sk, struct sk_buff *skb)
 902 {
 903         struct brnf_frag_data *data;
 904         int err;
 905
 906         data = this_cpu_ptr(&brnf_frag_data_storage);
 907         err = skb_cow_head(skb, data->size);
 908
 909         if (err) {
 910                 kfree_skb(skb);
 911                 return 0;
 912         }
 913
 914         if (data->vlan_tci) {
 915                 skb->vlan_tci = data->vlan_tci;
 916                 skb->vlan_proto = data->vlan_proto;
 917         }
 918
 919         skb_copy_to_linear_data_offset(skb, -data->size, data->mac, data->size);
 920         __skb_push(skb, data->encap_size);
 921
 922         nf_bridge_info_free(skb);
 923         return br_dev_queue_push_xmit(sk, skb);
 924 }
 925 #endif
 926
 927 static int br_nf_ip_fragment(struct sock *sk, struct sk_buff *skb,
 928                              int (*output)(struct sock *, struct sk_buff *))
 929 {
 930         unsigned int mtu = ip_skb_dst_mtu(skb);
 931         struct iphdr *iph = ip_hdr(skb);
 932         struct rtable *rt = skb_rtable(skb);
 933         struct net_device *dev = rt->dst.dev;
 934
 935         if (unlikely(((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) ||
 936                      (IPCB(skb)->frag_max_size &&
 937                       IPCB(skb)->frag_max_size > mtu))) {
 938                 IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
 939                 kfree_skb(skb);
 940                 return -EMSGSIZE;
 941         }
 942
 943         return ip_do_fragment(sk, skb, output);
 944 }
 945
 946 static unsigned int nf_bridge_mtu_reduction(const struct sk_buff *skb)
 947 {
 948         if (skb->nf_bridge->orig_proto == BRNF_PROTO_PPPOE)
 949                 return PPPOE_SES_HLEN;
 950         return 0;
 951 }
 952
 953 static int br_nf_dev_queue_xmit(struct sock *sk, struct sk_buff *skb)
 954 {
 955         struct nf_bridge_info *nf_bridge;
 956         unsigned int mtu_reserved;
 957
 958         mtu_reserved = nf_bridge_mtu_reduction(skb);
 959
 960         if (skb_is_gso(skb) || skb->len + mtu_reserved <= skb->dev->mtu) {
 961                 nf_bridge_info_free(skb);
 962                 return br_dev_queue_push_xmit(sk, skb);
 963         }
 964
 965         nf_bridge = nf_bridge_info_get(skb);
 966
 967 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4)
 968         /* This is wrong! We should preserve the original fragment
 969          * boundaries by preserving frag_list rather than refragmenting.
 970          */
 971         if (skb->protocol == htons(ETH_P_IP)) {
 972                 struct brnf_frag_data *data;
 973
 974                 if (br_validate_ipv4(skb))
 975                         return NF_DROP;
 976
 977                 IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
 978
 979                 nf_bridge_update_protocol(skb);
 980
 981                 data = this_cpu_ptr(&brnf_frag_data_storage);
 982
 983                 data->vlan_tci = skb->vlan_tci;
 984                 data->vlan_proto = skb->vlan_proto;
 985                 data->encap_size = nf_bridge_encap_header_len(skb);
 986                 data->size = ETH_HLEN + data->encap_size;
 987
 988                 skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
 989                                                  data->size);
 990
 991                 return br_nf_ip_fragment(sk, skb, br_nf_push_frag_xmit);
 992         }
 993 #endif
 994 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
 995         if (skb->protocol == htons(ETH_P_IPV6)) {
 996                 const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops();
 997                 struct brnf_frag_data *data;
 998
 999                 if (br_validate_ipv6(skb))
1000                         return NF_DROP;
1001
1002                 IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
1003
1004                 nf_bridge_update_protocol(skb);
1005
1006                 data = this_cpu_ptr(&brnf_frag_data_storage);
1007                 data->encap_size = nf_bridge_encap_header_len(skb);
1008                 data->size = ETH_HLEN + data->encap_size;
1009
1010                 skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
1011                                                  data->size);
1012
1013                 if (v6ops)
1014                         return v6ops->fragment(sk, skb, br_nf_push_frag_xmit);
1015                 else
1016                         return -EMSGSIZE;
1017         }
1018 #endif
1019         nf_bridge_info_free(skb);
1020         return br_dev_queue_push_xmit(sk, skb);
1021 }
1022
1023 /* PF_BRIDGE/POST_ROUTING ********************************************/
1024 static unsigned int br_nf_post_routing(const struct nf_hook_ops *ops,
1025                                        struct sk_buff *skb,
1026                                        const struct nf_hook_state *state)
1027 {
1028         struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
1029         struct net_device *realoutdev = bridge_parent(skb->dev);
1030         u_int8_t pf;
1031
1032         /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
1033          * on a bridge, but was delivered locally and is now being routed:
1034          *
1035          * POST_ROUTING was already invoked from the ip stack.
1036          */
1037         if (!nf_bridge || !nf_bridge->physoutdev)
1038                 return NF_ACCEPT;
1039
1040         if (!realoutdev)
1041                 return NF_DROP;
1042
1043         if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
1044                 pf = NFPROTO_IPV4;
1045         else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
1046                 pf = NFPROTO_IPV6;
1047         else
1048                 return NF_ACCEPT;
1049
1050         /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
1051          * about the value of skb->pkt_type. */
1052         if (skb->pkt_type == PACKET_OTHERHOST) {
1053                 skb->pkt_type = PACKET_HOST;
1054                 nf_bridge->pkt_otherhost = true;
1055         }
1056
1057         nf_bridge_pull_encap_header(skb);
1058         if (pf == NFPROTO_IPV4)
1059                 skb->protocol = htons(ETH_P_IP);
1060         else
1061                 skb->protocol = htons(ETH_P_IPV6);
1062
1063         NF_HOOK(pf, NF_INET_POST_ROUTING, state->sk, skb,
1064                 NULL, realoutdev,
1065                 br_nf_dev_queue_xmit);
1066
1067         return NF_STOLEN;
1068 }
1069
1070 /* IP/SABOTAGE *****************************************************/
1071 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
1072  * for the second time. */
1073 static unsigned int ip_sabotage_in(const struct nf_hook_ops *ops,
1074                                    struct sk_buff *skb,
1075                                    const struct nf_hook_state *state)
1076 {
1077         if (skb->nf_bridge &&
1078             !(skb->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)) {
1079                 return NF_STOP;
1080         }
1081
1082         return NF_ACCEPT;
1083 }
1084
1085 /* This is called when br_netfilter has called into iptables/netfilter,
1086  * and DNAT has taken place on a bridge-forwarded packet.
1087  *
1088  * neigh->output has created a new MAC header, with local br0 MAC
1089  * as saddr.
1090  *
1091  * This restores the original MAC saddr of the bridged packet
1092  * before invoking bridge forward logic to transmit the packet.
1093  */
1094 static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff *skb)
1095 {
1096         struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
1097
1098         skb_pull(skb, ETH_HLEN);
1099         nf_bridge->mask &= ~BRNF_BRIDGED_DNAT;
1100
1101         BUILD_BUG_ON(sizeof(nf_bridge->neigh_header) != (ETH_HLEN - ETH_ALEN));
1102
1103         skb_copy_to_linear_data_offset(skb, -(ETH_HLEN - ETH_ALEN),
1104                                        nf_bridge->neigh_header,
1105                                        ETH_HLEN - ETH_ALEN);
1106         skb->dev = nf_bridge->physindev;
1107
1108         nf_bridge->physoutdev = NULL;
1109         br_handle_frame_finish(NULL, skb);
1110 }
1111
1112 static int br_nf_dev_xmit(struct sk_buff *skb)
1113 {
1114         if (skb->nf_bridge && (skb->nf_bridge->mask & BRNF_BRIDGED_DNAT)) {
1115                 br_nf_pre_routing_finish_bridge_slow(skb);
1116                 return 1;
1117         }
1118         return 0;
1119 }
1120
1121 static const struct nf_br_ops br_ops = {
1122         .br_dev_xmit_hook =     br_nf_dev_xmit,
1123 };
1124
1125 void br_netfilter_enable(void)
1126 {
1127 }
1128 EXPORT_SYMBOL_GPL(br_netfilter_enable);
1129
1130 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
1131  * br_dev_queue_push_xmit is called afterwards */
1132 static struct nf_hook_ops br_nf_ops[] __read_mostly = {
1133         {
1134                 .hook = br_nf_pre_routing,
1135                 .owner = THIS_MODULE,
1136                 .pf = NFPROTO_BRIDGE,
1137                 .hooknum = NF_BR_PRE_ROUTING,
1138                 .priority = NF_BR_PRI_BRNF,
1139         },
1140         {
1141                 .hook = br_nf_local_in,
1142                 .owner = THIS_MODULE,
1143                 .pf = NFPROTO_BRIDGE,
1144                 .hooknum = NF_BR_LOCAL_IN,
1145                 .priority = NF_BR_PRI_BRNF,
1146         },
1147         {
1148                 .hook = br_nf_forward_ip,
1149                 .owner = THIS_MODULE,
1150                 .pf = NFPROTO_BRIDGE,
1151                 .hooknum = NF_BR_FORWARD,
1152                 .priority = NF_BR_PRI_BRNF - 1,
1153         },
1154         {
1155                 .hook = br_nf_forward_arp,
1156                 .owner = THIS_MODULE,
1157                 .pf = NFPROTO_BRIDGE,
1158                 .hooknum = NF_BR_FORWARD,
1159                 .priority = NF_BR_PRI_BRNF,
1160         },
1161         {
1162                 .hook = br_nf_post_routing,
1163                 .owner = THIS_MODULE,
1164                 .pf = NFPROTO_BRIDGE,
1165                 .hooknum = NF_BR_POST_ROUTING,
1166                 .priority = NF_BR_PRI_LAST,
1167         },
1168         {
1169                 .hook = ip_sabotage_in,
1170                 .owner = THIS_MODULE,
1171                 .pf = NFPROTO_IPV4,
1172                 .hooknum = NF_INET_PRE_ROUTING,
1173                 .priority = NF_IP_PRI_FIRST,
1174         },
1175         {
1176                 .hook = ip_sabotage_in,
1177                 .owner = THIS_MODULE,
1178                 .pf = NFPROTO_IPV6,
1179                 .hooknum = NF_INET_PRE_ROUTING,
1180                 .priority = NF_IP6_PRI_FIRST,
1181         },
1182 };
1183
1184 #ifdef CONFIG_SYSCTL
1185 static
1186 int brnf_sysctl_call_tables(struct ctl_table *ctl, int write,
1187                             void __user *buffer, size_t *lenp, loff_t *ppos)
1188 {
1189         int ret;
1190
1191         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1192
1193         if (write && *(int *)(ctl->data))
1194                 *(int *)(ctl->data) = 1;
1195         return ret;
1196 }
1197
1198 static struct ctl_table brnf_table[] = {
1199         {
1200                 .procname       = "bridge-nf-call-arptables",
1201                 .data           = &brnf_call_arptables,
1202                 .maxlen         = sizeof(int),
1203                 .mode           = 0644,
1204                 .proc_handler   = brnf_sysctl_call_tables,
1205         },
1206         {
1207                 .procname       = "bridge-nf-call-iptables",
1208                 .data           = &brnf_call_iptables,
1209                 .maxlen         = sizeof(int),
1210                 .mode           = 0644,
1211                 .proc_handler   = brnf_sysctl_call_tables,
1212         },
1213         {
1214                 .procname       = "bridge-nf-call-ip6tables",
1215                 .data           = &brnf_call_ip6tables,
1216                 .maxlen         = sizeof(int),
1217                 .mode           = 0644,
1218                 .proc_handler   = brnf_sysctl_call_tables,
1219         },
1220         {
1221                 .procname       = "bridge-nf-filter-vlan-tagged",
1222                 .data           = &brnf_filter_vlan_tagged,
1223                 .maxlen         = sizeof(int),
1224                 .mode           = 0644,
1225                 .proc_handler   = brnf_sysctl_call_tables,
1226         },
1227         {
1228                 .procname       = "bridge-nf-filter-pppoe-tagged",
1229                 .data           = &brnf_filter_pppoe_tagged,
1230                 .maxlen         = sizeof(int),
1231                 .mode           = 0644,
1232                 .proc_handler   = brnf_sysctl_call_tables,
1233         },
1234         {
1235                 .procname       = "bridge-nf-pass-vlan-input-dev",
1236                 .data           = &brnf_pass_vlan_indev,
1237                 .maxlen         = sizeof(int),
1238                 .mode           = 0644,
1239                 .proc_handler   = brnf_sysctl_call_tables,
1240         },
1241         { }
1242 };
1243 #endif
1244
1245 static int __init br_netfilter_init(void)
1246 {
1247         int ret;
1248
1249         ret = nf_register_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
1250         if (ret < 0)
1251                 return ret;
1252
1253 #ifdef CONFIG_SYSCTL
1254         brnf_sysctl_header = register_net_sysctl(&init_net, "net/bridge", brnf_table);
1255         if (brnf_sysctl_header == NULL) {
1256                 printk(KERN_WARNING
1257                        "br_netfilter: can't register to sysctl.\n");
1258                 nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
1259                 return -ENOMEM;
1260         }
1261 #endif
1262         RCU_INIT_POINTER(nf_br_ops, &br_ops);
1263         printk(KERN_NOTICE "Bridge firewalling registered\n");
1264         return 0;
1265 }
1266
1267 static void __exit br_netfilter_fini(void)
1268 {
1269         RCU_INIT_POINTER(nf_br_ops, NULL);
1270         nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
1271 #ifdef CONFIG_SYSCTL
1272         unregister_net_sysctl_table(brnf_sysctl_header);
1273 #endif
1274 }
1275
1276 module_init(br_netfilter_init);
1277 module_exit(br_netfilter_fini);
1278
1279 MODULE_LICENSE("GPL");
1280 MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
1281 MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
1282 MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");