net/bridge/br_netfilter.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/route.h>
  38
  39 #include <asm/uaccess.h>
  40 #include "br_private.h"
  41 #ifdef CONFIG_SYSCTL
  42 #include <linux/sysctl.h>
  43 #endif
  44
  45 #define skb_origaddr(skb)        (((struct bridge_skb_cb *) \
  46                                  (skb->nf_bridge->data))->daddr.ipv4)
  47 #define store_orig_dstaddr(skb)  (skb_origaddr(skb) = ip_hdr(skb)->daddr)
  48 #define dnat_took_place(skb)     (skb_origaddr(skb) != ip_hdr(skb)->daddr)
  49
  50 #ifdef CONFIG_SYSCTL
  51 static struct ctl_table_header *brnf_sysctl_header;
  52 static int brnf_call_iptables __read_mostly = 1;
  53 static int brnf_call_ip6tables __read_mostly = 1;
  54 static int brnf_call_arptables __read_mostly = 1;
  55 static int brnf_filter_vlan_tagged __read_mostly = 0;
  56 static int brnf_filter_pppoe_tagged __read_mostly = 0;
  57 #else
  58 #define brnf_call_iptables 1
  59 #define brnf_call_ip6tables 1
  60 #define brnf_call_arptables 1
  61 #define brnf_filter_vlan_tagged 0
  62 #define brnf_filter_pppoe_tagged 0
  63 #endif
  64
  65 static inline __be16 vlan_proto(const struct sk_buff *skb)
  66 {
  67         return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
  68 }
  69
  70 #define IS_VLAN_IP(skb) \
  71         (skb->protocol == htons(ETH_P_8021Q) && \
  72          vlan_proto(skb) == htons(ETH_P_IP) &&  \
  73          brnf_filter_vlan_tagged)
  74
  75 #define IS_VLAN_IPV6(skb) \
  76         (skb->protocol == htons(ETH_P_8021Q) && \
  77          vlan_proto(skb) == htons(ETH_P_IPV6) &&\
  78          brnf_filter_vlan_tagged)
  79
  80 #define IS_VLAN_ARP(skb) \
  81         (skb->protocol == htons(ETH_P_8021Q) && \
  82          vlan_proto(skb) == htons(ETH_P_ARP) && \
  83          brnf_filter_vlan_tagged)
  84
  85 static inline __be16 pppoe_proto(const struct sk_buff *skb)
  86 {
  87         return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
  88                             sizeof(struct pppoe_hdr)));
  89 }
  90
  91 #define IS_PPPOE_IP(skb) \
  92         (skb->protocol == htons(ETH_P_PPP_SES) && \
  93          pppoe_proto(skb) == htons(PPP_IP) && \
  94          brnf_filter_pppoe_tagged)
  95
  96 #define IS_PPPOE_IPV6(skb) \
  97         (skb->protocol == htons(ETH_P_PPP_SES) && \
  98          pppoe_proto(skb) == htons(PPP_IPV6) && \
  99          brnf_filter_pppoe_tagged)
 100
 101 static void fake_update_pmtu(struct dst_entry *dst, u32 mtu)
 102 {
 103 }
 104
 105 static struct dst_ops fake_dst_ops = {
 106         .family =               AF_INET,
 107         .protocol =             cpu_to_be16(ETH_P_IP),
 108         .update_pmtu =          fake_update_pmtu,
 109         .entries =              ATOMIC_INIT(0),
 110 };
 111
 112 /*
 113  * Initialize bogus route table used to keep netfilter happy.
 114  * Currently, we fill in the PMTU entry because netfilter
 115  * refragmentation needs it, and the rt_flags entry because
 116  * ipt_REJECT needs it.  Future netfilter modules might
 117  * require us to fill additional fields.
 118  */
 119 void br_netfilter_rtable_init(struct net_bridge *br)
 120 {
 121         struct rtable *rt = &br->fake_rtable;
 122
 123         atomic_set(&rt->dst.__refcnt, 1);
 124         rt->dst.dev = br->dev;
 125         rt->dst.path = &rt->dst;
 126         rt->dst.metrics[RTAX_MTU - 1] = 1500;
 127         rt->dst.flags   = DST_NOXFRM;
 128         rt->dst.ops = &fake_dst_ops;
 129 }
 130
 131 static inline struct rtable *bridge_parent_rtable(const struct net_device *dev)
 132 {
 133         if (!br_port_exists(dev))
 134                 return NULL;
 135         return &br_port_get_rcu(dev)->br->fake_rtable;
 136 }
 137
 138 static inline struct net_device *bridge_parent(const struct net_device *dev)
 139 {
 140         if (!br_port_exists(dev))
 141                 return NULL;
 142
 143         return br_port_get_rcu(dev)->br->dev;
 144 }
 145
 146 static inline struct nf_bridge_info *nf_bridge_alloc(struct sk_buff *skb)
 147 {
 148         skb->nf_bridge = kzalloc(sizeof(struct nf_bridge_info), GFP_ATOMIC);
 149         if (likely(skb->nf_bridge))
 150                 atomic_set(&(skb->nf_bridge->use), 1);
 151
 152         return skb->nf_bridge;
 153 }
 154
 155 static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
 156 {
 157         struct nf_bridge_info *nf_bridge = skb->nf_bridge;
 158
 159         if (atomic_read(&nf_bridge->use) > 1) {
 160                 struct nf_bridge_info *tmp = nf_bridge_alloc(skb);
 161
 162                 if (tmp) {
 163                         memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info));
 164                         atomic_set(&tmp->use, 1);
 165                         nf_bridge_put(nf_bridge);
 166                 }
 167                 nf_bridge = tmp;
 168         }
 169         return nf_bridge;
 170 }
 171
 172 static inline void nf_bridge_push_encap_header(struct sk_buff *skb)
 173 {
 174         unsigned int len = nf_bridge_encap_header_len(skb);
 175
 176         skb_push(skb, len);
 177         skb->network_header -= len;
 178 }
 179
 180 static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
 181 {
 182         unsigned int len = nf_bridge_encap_header_len(skb);
 183
 184         skb_pull(skb, len);
 185         skb->network_header += len;
 186 }
 187
 188 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
 189 {
 190         unsigned int len = nf_bridge_encap_header_len(skb);
 191
 192         skb_pull_rcsum(skb, len);
 193         skb->network_header += len;
 194 }
 195
 196 static inline void nf_bridge_save_header(struct sk_buff *skb)
 197 {
 198         int header_size = ETH_HLEN + nf_bridge_encap_header_len(skb);
 199
 200         skb_copy_from_linear_data_offset(skb, -header_size,
 201                                          skb->nf_bridge->data, header_size);
 202 }
 203
 204 static inline void nf_bridge_update_protocol(struct sk_buff *skb)
 205 {
 206         if (skb->nf_bridge->mask & BRNF_8021Q)
 207                 skb->protocol = htons(ETH_P_8021Q);
 208         else if (skb->nf_bridge->mask & BRNF_PPPoE)
 209                 skb->protocol = htons(ETH_P_PPP_SES);
 210 }
 211
 212 /* Fill in the header for fragmented IP packets handled by
 213  * the IPv4 connection tracking code.
 214  */
 215 int nf_bridge_copy_header(struct sk_buff *skb)
 216 {
 217         int err;
 218         unsigned int header_size;
 219
 220         nf_bridge_update_protocol(skb);
 221         header_size = ETH_HLEN + nf_bridge_encap_header_len(skb);
 222         err = skb_cow_head(skb, header_size);
 223         if (err)
 224                 return err;
 225
 226         skb_copy_to_linear_data_offset(skb, -header_size,
 227                                        skb->nf_bridge->data, header_size);
 228         __skb_push(skb, nf_bridge_encap_header_len(skb));
 229         return 0;
 230 }
 231
 232 /* PF_BRIDGE/PRE_ROUTING *********************************************/
 233 /* Undo the changes made for ip6tables PREROUTING and continue the
 234  * bridge PRE_ROUTING hook. */
 235 static int br_nf_pre_routing_finish_ipv6(struct sk_buff *skb)
 236 {
 237         struct nf_bridge_info *nf_bridge = skb->nf_bridge;
 238         struct rtable *rt;
 239
 240         if (nf_bridge->mask & BRNF_PKT_TYPE) {
 241                 skb->pkt_type = PACKET_OTHERHOST;
 242                 nf_bridge->mask ^= BRNF_PKT_TYPE;
 243         }
 244         nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
 245
 246         rt = bridge_parent_rtable(nf_bridge->physindev);
 247         if (!rt) {
 248                 kfree_skb(skb);
 249                 return 0;
 250         }
 251         skb_dst_set_noref(skb, &rt->dst);
 252
 253         skb->dev = nf_bridge->physindev;
 254         nf_bridge_update_protocol(skb);
 255         nf_bridge_push_encap_header(skb);
 256         NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
 257                        br_handle_frame_finish, 1);
 258
 259         return 0;
 260 }
 261
 262 /* Obtain the correct destination MAC address, while preserving the original
 263  * source MAC address. If we already know this address, we just copy it. If we
 264  * don't, we use the neighbour framework to find out. In both cases, we make
 265  * sure that br_handle_frame_finish() is called afterwards.
 266  */
 267 static int br_nf_pre_routing_finish_bridge(struct sk_buff *skb)
 268 {
 269         struct nf_bridge_info *nf_bridge = skb->nf_bridge;
 270         struct dst_entry *dst;
 271
 272         skb->dev = bridge_parent(skb->dev);
 273         if (!skb->dev)
 274                 goto free_skb;
 275         dst = skb_dst(skb);
 276         if (dst->hh) {
 277                 neigh_hh_bridge(dst->hh, skb);
 278                 skb->dev = nf_bridge->physindev;
 279                 return br_handle_frame_finish(skb);
 280         } else if (dst->neighbour) {
 281                 /* the neighbour function below overwrites the complete
 282                  * MAC header, so we save the Ethernet source address and
 283                  * protocol number. */
 284                 skb_copy_from_linear_data_offset(skb, -(ETH_HLEN-ETH_ALEN), skb->nf_bridge->data, ETH_HLEN-ETH_ALEN);
 285                 /* tell br_dev_xmit to continue with forwarding */
 286                 nf_bridge->mask |= BRNF_BRIDGED_DNAT;
 287                 return dst->neighbour->output(skb);
 288         }
 289 free_skb:
 290         kfree_skb(skb);
 291         return 0;
 292 }
 293
 294 /* This requires some explaining. If DNAT has taken place,
 295  * we will need to fix up the destination Ethernet address.
 296  *
 297  * There are two cases to consider:
 298  * 1. The packet was DNAT'ed to a device in the same bridge
 299  *    port group as it was received on. We can still bridge
 300  *    the packet.
 301  * 2. The packet was DNAT'ed to a different device, either
 302  *    a non-bridged device or another bridge port group.
 303  *    The packet will need to be routed.
 304  *
 305  * The correct way of distinguishing between these two cases is to
 306  * call ip_route_input() and to look at skb->dst->dev, which is
 307  * changed to the destination device if ip_route_input() succeeds.
 308  *
 309  * Let's first consider the case that ip_route_input() succeeds:
 310  *
 311  * If the output device equals the logical bridge device the packet
 312  * came in on, we can consider this bridging. The corresponding MAC
 313  * address will be obtained in br_nf_pre_routing_finish_bridge.
 314  * Otherwise, the packet is considered to be routed and we just
 315  * change the destination MAC address so that the packet will
 316  * later be passed up to the IP stack to be routed. For a redirected
 317  * packet, ip_route_input() will give back the localhost as output device,
 318  * which differs from the bridge device.
 319  *
 320  * Let's now consider the case that ip_route_input() fails:
 321  *
 322  * This can be because the destination address is martian, in which case
 323  * the packet will be dropped.
 324  * If IP forwarding is disabled, ip_route_input() will fail, while
 325  * ip_route_output_key() can return success. The source
 326  * address for ip_route_output_key() is set to zero, so ip_route_output_key()
 327  * thinks we're handling a locally generated packet and won't care
 328  * if IP forwarding is enabled. If the output device equals the logical bridge
 329  * device, we proceed as if ip_route_input() succeeded. If it differs from the
 330  * logical bridge port or if ip_route_output_key() fails we drop the packet.
 331  */
 332 static int br_nf_pre_routing_finish(struct sk_buff *skb)
 333 {
 334         struct net_device *dev = skb->dev;
 335         struct iphdr *iph = ip_hdr(skb);
 336         struct nf_bridge_info *nf_bridge = skb->nf_bridge;
 337         struct rtable *rt;
 338         int err;
 339
 340         if (nf_bridge->mask & BRNF_PKT_TYPE) {
 341                 skb->pkt_type = PACKET_OTHERHOST;
 342                 nf_bridge->mask ^= BRNF_PKT_TYPE;
 343         }
 344         nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
 345         if (dnat_took_place(skb)) {
 346                 if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
 347                         struct flowi fl = {
 348                                 .nl_u = {
 349                                         .ip4_u = {
 350                                                  .daddr = iph->daddr,
 351                                                  .saddr = 0,
 352                                                  .tos = RT_TOS(iph->tos) },
 353                                 },
 354                                 .proto = 0,
 355                         };
 356                         struct in_device *in_dev = __in_dev_get_rcu(dev);
 357
 358                         /* If err equals -EHOSTUNREACH the error is due to a
 359                          * martian destination or due to the fact that
 360                          * forwarding is disabled. For most martian packets,
 361                          * ip_route_output_key() will fail. It won't fail for 2 types of
 362                          * martian destinations: loopback destinations and destination
 363                          * 0.0.0.0. In both cases the packet will be dropped because the
 364                          * destination is the loopback device and not the bridge. */
 365                         if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
 366                                 goto free_skb;
 367
 368                         if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
 369                                 /* - Bridged-and-DNAT'ed traffic doesn't
 370                                  *   require ip_forwarding. */
 371                                 if (((struct dst_entry *)rt)->dev == dev) {
 372                                         skb_dst_set(skb, (struct dst_entry *)rt);
 373                                         goto bridged_dnat;
 374                                 }
 375                                 dst_release((struct dst_entry *)rt);
 376                         }
 377 free_skb:
 378                         kfree_skb(skb);
 379                         return 0;
 380                 } else {
 381                         if (skb_dst(skb)->dev == dev) {
 382 bridged_dnat:
 383                                 skb->dev = nf_bridge->physindev;
 384                                 nf_bridge_update_protocol(skb);
 385                                 nf_bridge_push_encap_header(skb);
 386                                 NF_HOOK_THRESH(NFPROTO_BRIDGE,
 387                                                NF_BR_PRE_ROUTING,
 388                                                skb, skb->dev, NULL,
 389                                                br_nf_pre_routing_finish_bridge,
 390                                                1);
 391                                 return 0;
 392                         }
 393                         memcpy(eth_hdr(skb)->h_dest, dev->dev_addr, ETH_ALEN);
 394                         skb->pkt_type = PACKET_HOST;
 395                 }
 396         } else {
 397                 rt = bridge_parent_rtable(nf_bridge->physindev);
 398                 if (!rt) {
 399                         kfree_skb(skb);
 400                         return 0;
 401                 }
 402                 skb_dst_set_noref(skb, &rt->dst);
 403         }
 404
 405         skb->dev = nf_bridge->physindev;
 406         nf_bridge_update_protocol(skb);
 407         nf_bridge_push_encap_header(skb);
 408         NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
 409                        br_handle_frame_finish, 1);
 410
 411         return 0;
 412 }
 413
 414 /* Some common code for IPv4/IPv6 */
 415 static struct net_device *setup_pre_routing(struct sk_buff *skb)
 416 {
 417         struct nf_bridge_info *nf_bridge = skb->nf_bridge;
 418
 419         if (skb->pkt_type == PACKET_OTHERHOST) {
 420                 skb->pkt_type = PACKET_HOST;
 421                 nf_bridge->mask |= BRNF_PKT_TYPE;
 422         }
 423
 424         nf_bridge->mask |= BRNF_NF_BRIDGE_PREROUTING;
 425         nf_bridge->physindev = skb->dev;
 426         skb->dev = bridge_parent(skb->dev);
 427         if (skb->protocol == htons(ETH_P_8021Q))
 428                 nf_bridge->mask |= BRNF_8021Q;
 429         else if (skb->protocol == htons(ETH_P_PPP_SES))
 430                 nf_bridge->mask |= BRNF_PPPoE;
 431
 432         return skb->dev;
 433 }
 434
 435 /* We only check the length. A bridge shouldn't do any hop-by-hop stuff anyway */
 436 static int check_hbh_len(struct sk_buff *skb)
 437 {
 438         unsigned char *raw = (u8 *)(ipv6_hdr(skb) + 1);
 439         u32 pkt_len;
 440         const unsigned char *nh = skb_network_header(skb);
 441         int off = raw - nh;
 442         int len = (raw[1] + 1) << 3;
 443
 444         if ((raw + len) - skb->data > skb_headlen(skb))
 445                 goto bad;
 446
 447         off += 2;
 448         len -= 2;
 449
 450         while (len > 0) {
 451                 int optlen = nh[off + 1] + 2;
 452
 453                 switch (nh[off]) {
 454                 case IPV6_TLV_PAD0:
 455                         optlen = 1;
 456                         break;
 457
 458                 case IPV6_TLV_PADN:
 459                         break;
 460
 461                 case IPV6_TLV_JUMBO:
 462                         if (nh[off + 1] != 4 || (off & 3) != 2)
 463                                 goto bad;
 464                         pkt_len = ntohl(*(__be32 *) (nh + off + 2));
 465                         if (pkt_len <= IPV6_MAXPLEN ||
 466                             ipv6_hdr(skb)->payload_len)
 467                                 goto bad;
 468                         if (pkt_len > skb->len - sizeof(struct ipv6hdr))
 469                                 goto bad;
 470                         if (pskb_trim_rcsum(skb,
 471                                             pkt_len + sizeof(struct ipv6hdr)))
 472                                 goto bad;
 473                         nh = skb_network_header(skb);
 474                         break;
 475                 default:
 476                         if (optlen > len)
 477                                 goto bad;
 478                         break;
 479                 }
 480                 off += optlen;
 481                 len -= optlen;
 482         }
 483         if (len == 0)
 484                 return 0;
 485 bad:
 486         return -1;
 487
 488 }
 489
 490 /* Replicate the checks that IPv6 does on packet reception and pass the packet
 491  * to ip6tables, which doesn't support NAT, so things are fairly simple. */
 492 static unsigned int br_nf_pre_routing_ipv6(unsigned int hook,
 493                                            struct sk_buff *skb,
 494                                            const struct net_device *in,
 495                                            const struct net_device *out,
 496                                            int (*okfn)(struct sk_buff *))
 497 {
 498         struct ipv6hdr *hdr;
 499         u32 pkt_len;
 500
 501         if (skb->len < sizeof(struct ipv6hdr))
 502                 goto inhdr_error;
 503
 504         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 505                 goto inhdr_error;
 506
 507         hdr = ipv6_hdr(skb);
 508
 509         if (hdr->version != 6)
 510                 goto inhdr_error;
 511
 512         pkt_len = ntohs(hdr->payload_len);
 513
 514         if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
 515                 if (pkt_len + sizeof(struct ipv6hdr) > skb->len)
 516                         goto inhdr_error;
 517                 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
 518                         goto inhdr_error;
 519         }
 520         if (hdr->nexthdr == NEXTHDR_HOP && check_hbh_len(skb))
 521                 goto inhdr_error;
 522
 523         nf_bridge_put(skb->nf_bridge);
 524         if (!nf_bridge_alloc(skb))
 525                 return NF_DROP;
 526         if (!setup_pre_routing(skb))
 527                 return NF_DROP;
 528
 529         skb->protocol = htons(ETH_P_IPV6);
 530         NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
 531                 br_nf_pre_routing_finish_ipv6);
 532
 533         return NF_STOLEN;
 534
 535 inhdr_error:
 536         return NF_DROP;
 537 }
 538
 539 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
 540  * Replicate the checks that IPv4 does on packet reception.
 541  * Set skb->dev to the bridge device (i.e. parent of the
 542  * receiving device) to make netfilter happy, the REDIRECT
 543  * target in particular.  Save the original destination IP
 544  * address to be able to detect DNAT afterwards. */
 545 static unsigned int br_nf_pre_routing(unsigned int hook, struct sk_buff *skb,
 546                                       const struct net_device *in,
 547                                       const struct net_device *out,
 548                                       int (*okfn)(struct sk_buff *))
 549 {
 550         struct net_bridge_port *p;
 551         struct net_bridge *br;
 552         struct iphdr *iph;
 553         __u32 len = nf_bridge_encap_header_len(skb);
 554
 555         if (unlikely(!pskb_may_pull(skb, len)))
 556                 goto out;
 557
 558         p = br_port_get_rcu(in);
 559         if (p == NULL)
 560                 goto out;
 561         br = p->br;
 562
 563         if (skb->protocol == htons(ETH_P_IPV6) || IS_VLAN_IPV6(skb) ||
 564             IS_PPPOE_IPV6(skb)) {
 565                 if (!brnf_call_ip6tables && !br->nf_call_ip6tables)
 566                         return NF_ACCEPT;
 567
 568                 nf_bridge_pull_encap_header_rcsum(skb);
 569                 return br_nf_pre_routing_ipv6(hook, skb, in, out, okfn);
 570         }
 571
 572         if (!brnf_call_iptables && !br->nf_call_iptables)
 573                 return NF_ACCEPT;
 574
 575         if (skb->protocol != htons(ETH_P_IP) && !IS_VLAN_IP(skb) &&
 576             !IS_PPPOE_IP(skb))
 577                 return NF_ACCEPT;
 578
 579         nf_bridge_pull_encap_header_rcsum(skb);
 580
 581         if (!pskb_may_pull(skb, sizeof(struct iphdr)))
 582                 goto inhdr_error;
 583
 584         iph = ip_hdr(skb);
 585         if (iph->ihl < 5 || iph->version != 4)
 586                 goto inhdr_error;
 587
 588         if (!pskb_may_pull(skb, 4 * iph->ihl))
 589                 goto inhdr_error;
 590
 591         iph = ip_hdr(skb);
 592         if (ip_fast_csum((__u8 *) iph, iph->ihl) != 0)
 593                 goto inhdr_error;
 594
 595         len = ntohs(iph->tot_len);
 596         if (skb->len < len || len < 4 * iph->ihl)
 597                 goto inhdr_error;
 598
 599         pskb_trim_rcsum(skb, len);
 600
 601         /* BUG: Should really parse the IP options here. */
 602         memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
 603
 604         nf_bridge_put(skb->nf_bridge);
 605         if (!nf_bridge_alloc(skb))
 606                 return NF_DROP;
 607         if (!setup_pre_routing(skb))
 608                 return NF_DROP;
 609         store_orig_dstaddr(skb);
 610         skb->protocol = htons(ETH_P_IP);
 611
 612         NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
 613                 br_nf_pre_routing_finish);
 614
 615         return NF_STOLEN;
 616
 617 inhdr_error:
 618 //      IP_INC_STATS_BH(IpInHdrErrors);
 619 out:
 620         return NF_DROP;
 621 }
 622
 623
 624 /* PF_BRIDGE/LOCAL_IN ************************************************/
 625 /* The packet is locally destined, which requires a real
 626  * dst_entry, so detach the fake one.  On the way up, the
 627  * packet would pass through PRE_ROUTING again (which already
 628  * took place when the packet entered the bridge), but we
 629  * register an IPv4 PRE_ROUTING 'sabotage' hook that will
 630  * prevent this from happening. */
 631 static unsigned int br_nf_local_in(unsigned int hook, struct sk_buff *skb,
 632                                    const struct net_device *in,
 633                                    const struct net_device *out,
 634                                    int (*okfn)(struct sk_buff *))
 635 {
 636         struct rtable *rt = skb_rtable(skb);
 637
 638         if (rt && rt == bridge_parent_rtable(in))
 639                 skb_dst_drop(skb);
 640
 641         return NF_ACCEPT;
 642 }
 643
 644 /* PF_BRIDGE/FORWARD *************************************************/
 645 static int br_nf_forward_finish(struct sk_buff *skb)
 646 {
 647         struct nf_bridge_info *nf_bridge = skb->nf_bridge;
 648         struct net_device *in;
 649
 650         if (skb->protocol != htons(ETH_P_ARP) && !IS_VLAN_ARP(skb)) {
 651                 in = nf_bridge->physindev;
 652                 if (nf_bridge->mask & BRNF_PKT_TYPE) {
 653                         skb->pkt_type = PACKET_OTHERHOST;
 654                         nf_bridge->mask ^= BRNF_PKT_TYPE;
 655                 }
 656                 nf_bridge_update_protocol(skb);
 657         } else {
 658                 in = *((struct net_device **)(skb->cb));
 659         }
 660         nf_bridge_push_encap_header(skb);
 661
 662         NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_FORWARD, skb, in,
 663                        skb->dev, br_forward_finish, 1);
 664         return 0;
 665 }
 666
 667 /* This is the 'purely bridged' case.  For IP, we pass the packet to
 668  * netfilter with indev and outdev set to the bridge device,
 669  * but we are still able to filter on the 'real' indev/outdev
 670  * because of the physdev module. For ARP, indev and outdev are the
 671  * bridge ports. */
 672 static unsigned int br_nf_forward_ip(unsigned int hook, struct sk_buff *skb,
 673                                      const struct net_device *in,
 674                                      const struct net_device *out,
 675                                      int (*okfn)(struct sk_buff *))
 676 {
 677         struct nf_bridge_info *nf_bridge;
 678         struct net_device *parent;
 679         u_int8_t pf;
 680
 681         if (!skb->nf_bridge)
 682                 return NF_ACCEPT;
 683
 684         /* Need exclusive nf_bridge_info since we might have multiple
 685          * different physoutdevs. */
 686         if (!nf_bridge_unshare(skb))
 687                 return NF_DROP;
 688
 689         parent = bridge_parent(out);
 690         if (!parent)
 691                 return NF_DROP;
 692
 693         if (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb) ||
 694             IS_PPPOE_IP(skb))
 695                 pf = PF_INET;
 696         else if (skb->protocol == htons(ETH_P_IPV6) || IS_VLAN_IPV6(skb) ||
 697                  IS_PPPOE_IPV6(skb))
 698                 pf = PF_INET6;
 699         else
 700                 return NF_ACCEPT;
 701
 702         nf_bridge_pull_encap_header(skb);
 703
 704         nf_bridge = skb->nf_bridge;
 705         if (skb->pkt_type == PACKET_OTHERHOST) {
 706                 skb->pkt_type = PACKET_HOST;
 707                 nf_bridge->mask |= BRNF_PKT_TYPE;
 708         }
 709
 710         /* The physdev module checks on this */
 711         nf_bridge->mask |= BRNF_BRIDGED;
 712         nf_bridge->physoutdev = skb->dev;
 713         if (pf == PF_INET)
 714                 skb->protocol = htons(ETH_P_IP);
 715         else
 716                 skb->protocol = htons(ETH_P_IPV6);
 717
 718         NF_HOOK(pf, NF_INET_FORWARD, skb, bridge_parent(in), parent,
 719                 br_nf_forward_finish);
 720
 721         return NF_STOLEN;
 722 }
 723
 724 static unsigned int br_nf_forward_arp(unsigned int hook, struct sk_buff *skb,
 725                                       const struct net_device *in,
 726                                       const struct net_device *out,
 727                                       int (*okfn)(struct sk_buff *))
 728 {
 729         struct net_bridge_port *p;
 730         struct net_bridge *br;
 731         struct net_device **d = (struct net_device **)(skb->cb);
 732
 733         p = br_port_get_rcu(out);
 734         if (p == NULL)
 735                 return NF_ACCEPT;
 736         br = p->br;
 737
 738         if (!brnf_call_arptables && !br->nf_call_arptables)
 739                 return NF_ACCEPT;
 740
 741         if (skb->protocol != htons(ETH_P_ARP)) {
 742                 if (!IS_VLAN_ARP(skb))
 743                         return NF_ACCEPT;
 744                 nf_bridge_pull_encap_header(skb);
 745         }
 746
 747         if (arp_hdr(skb)->ar_pln != 4) {
 748                 if (IS_VLAN_ARP(skb))
 749                         nf_bridge_push_encap_header(skb);
 750                 return NF_ACCEPT;
 751         }
 752         *d = (struct net_device *)in;
 753         NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, skb, (struct net_device *)in,
 754                 (struct net_device *)out, br_nf_forward_finish);
 755
 756         return NF_STOLEN;
 757 }
 758
 759 #if defined(CONFIG_NF_CONNTRACK_IPV4) || defined(CONFIG_NF_CONNTRACK_IPV4_MODULE)
 760 static int br_nf_dev_queue_xmit(struct sk_buff *skb)
 761 {
 762         if (skb->nfct != NULL && skb->protocol == htons(ETH_P_IP) &&
 763             skb->len + nf_bridge_mtu_reduction(skb) > skb->dev->mtu &&
 764             !skb_is_gso(skb))
 765                 return ip_fragment(skb, br_dev_queue_push_xmit);
 766         else
 767                 return br_dev_queue_push_xmit(skb);
 768 }
 769 #else
 770 static int br_nf_dev_queue_xmit(struct sk_buff *skb)
 771 {
 772         return br_dev_queue_push_xmit(skb);
 773 }
 774 #endif
 775
 776 /* PF_BRIDGE/POST_ROUTING ********************************************/
 777 static unsigned int br_nf_post_routing(unsigned int hook, struct sk_buff *skb,
 778                                        const struct net_device *in,
 779                                        const struct net_device *out,
 780                                        int (*okfn)(struct sk_buff *))
 781 {
 782         struct nf_bridge_info *nf_bridge = skb->nf_bridge;
 783         struct net_device *realoutdev = bridge_parent(skb->dev);
 784         u_int8_t pf;
 785
 786         if (!nf_bridge || !(nf_bridge->mask & BRNF_BRIDGED))
 787                 return NF_ACCEPT;
 788
 789         if (!realoutdev)
 790                 return NF_DROP;
 791
 792         if (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb) ||
 793             IS_PPPOE_IP(skb))
 794                 pf = PF_INET;
 795         else if (skb->protocol == htons(ETH_P_IPV6) || IS_VLAN_IPV6(skb) ||
 796                  IS_PPPOE_IPV6(skb))
 797                 pf = PF_INET6;
 798         else
 799                 return NF_ACCEPT;
 800
 801         /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
 802          * about the value of skb->pkt_type. */
 803         if (skb->pkt_type == PACKET_OTHERHOST) {
 804                 skb->pkt_type = PACKET_HOST;
 805                 nf_bridge->mask |= BRNF_PKT_TYPE;
 806         }
 807
 808         nf_bridge_pull_encap_header(skb);
 809         nf_bridge_save_header(skb);
 810         if (pf == PF_INET)
 811                 skb->protocol = htons(ETH_P_IP);
 812         else
 813                 skb->protocol = htons(ETH_P_IPV6);
 814
 815         NF_HOOK(pf, NF_INET_POST_ROUTING, skb, NULL, realoutdev,
 816                 br_nf_dev_queue_xmit);
 817
 818         return NF_STOLEN;
 819 }
 820
 821 /* IP/SABOTAGE *****************************************************/
 822 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
 823  * for the second time. */
 824 static unsigned int ip_sabotage_in(unsigned int hook, struct sk_buff *skb,
 825                                    const struct net_device *in,
 826                                    const struct net_device *out,
 827                                    int (*okfn)(struct sk_buff *))
 828 {
 829         if (skb->nf_bridge &&
 830             !(skb->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)) {
 831                 return NF_STOP;
 832         }
 833
 834         return NF_ACCEPT;
 835 }
 836
 837 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
 838  * br_dev_queue_push_xmit is called afterwards */
 839 static struct nf_hook_ops br_nf_ops[] __read_mostly = {
 840         {
 841                 .hook = br_nf_pre_routing,
 842                 .owner = THIS_MODULE,
 843                 .pf = PF_BRIDGE,
 844                 .hooknum = NF_BR_PRE_ROUTING,
 845                 .priority = NF_BR_PRI_BRNF,
 846         },
 847         {
 848                 .hook = br_nf_local_in,
 849                 .owner = THIS_MODULE,
 850                 .pf = PF_BRIDGE,
 851                 .hooknum = NF_BR_LOCAL_IN,
 852                 .priority = NF_BR_PRI_BRNF,
 853         },
 854         {
 855                 .hook = br_nf_forward_ip,
 856                 .owner = THIS_MODULE,
 857                 .pf = PF_BRIDGE,
 858                 .hooknum = NF_BR_FORWARD,
 859                 .priority = NF_BR_PRI_BRNF - 1,
 860         },
 861         {
 862                 .hook = br_nf_forward_arp,
 863                 .owner = THIS_MODULE,
 864                 .pf = PF_BRIDGE,
 865                 .hooknum = NF_BR_FORWARD,
 866                 .priority = NF_BR_PRI_BRNF,
 867         },
 868         {
 869                 .hook = br_nf_post_routing,
 870                 .owner = THIS_MODULE,
 871                 .pf = PF_BRIDGE,
 872                 .hooknum = NF_BR_POST_ROUTING,
 873                 .priority = NF_BR_PRI_LAST,
 874         },
 875         {
 876                 .hook = ip_sabotage_in,
 877                 .owner = THIS_MODULE,
 878                 .pf = PF_INET,
 879                 .hooknum = NF_INET_PRE_ROUTING,
 880                 .priority = NF_IP_PRI_FIRST,
 881         },
 882         {
 883                 .hook = ip_sabotage_in,
 884                 .owner = THIS_MODULE,
 885                 .pf = PF_INET6,
 886                 .hooknum = NF_INET_PRE_ROUTING,
 887                 .priority = NF_IP6_PRI_FIRST,
 888         },
 889 };
 890
 891 #ifdef CONFIG_SYSCTL
 892 static
 893 int brnf_sysctl_call_tables(ctl_table * ctl, int write,
 894                             void __user * buffer, size_t * lenp, loff_t * ppos)
 895 {
 896         int ret;
 897
 898         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
 899
 900         if (write && *(int *)(ctl->data))
 901                 *(int *)(ctl->data) = 1;
 902         return ret;
 903 }
 904
 905 static ctl_table brnf_table[] = {
 906         {
 907                 .procname       = "bridge-nf-call-arptables",
 908                 .data           = &brnf_call_arptables,
 909                 .maxlen         = sizeof(int),
 910                 .mode           = 0644,
 911                 .proc_handler   = brnf_sysctl_call_tables,
 912         },
 913         {
 914                 .procname       = "bridge-nf-call-iptables",
 915                 .data           = &brnf_call_iptables,
 916                 .maxlen         = sizeof(int),
 917                 .mode           = 0644,
 918                 .proc_handler   = brnf_sysctl_call_tables,
 919         },
 920         {
 921                 .procname       = "bridge-nf-call-ip6tables",
 922                 .data           = &brnf_call_ip6tables,
 923                 .maxlen         = sizeof(int),
 924                 .mode           = 0644,
 925                 .proc_handler   = brnf_sysctl_call_tables,
 926         },
 927         {
 928                 .procname       = "bridge-nf-filter-vlan-tagged",
 929                 .data           = &brnf_filter_vlan_tagged,
 930                 .maxlen         = sizeof(int),
 931                 .mode           = 0644,
 932                 .proc_handler   = brnf_sysctl_call_tables,
 933         },
 934         {
 935                 .procname       = "bridge-nf-filter-pppoe-tagged",
 936                 .data           = &brnf_filter_pppoe_tagged,
 937                 .maxlen         = sizeof(int),
 938                 .mode           = 0644,
 939                 .proc_handler   = brnf_sysctl_call_tables,
 940         },
 941         { }
 942 };
 943
 944 static struct ctl_path brnf_path[] = {
 945         { .procname = "net", },
 946         { .procname = "bridge", },
 947         { }
 948 };
 949 #endif
 950
 951 int __init br_netfilter_init(void)
 952 {
 953         int ret;
 954
 955         ret = nf_register_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
 956         if (ret < 0)
 957                 return ret;
 958 #ifdef CONFIG_SYSCTL
 959         brnf_sysctl_header = register_sysctl_paths(brnf_path, brnf_table);
 960         if (brnf_sysctl_header == NULL) {
 961                 printk(KERN_WARNING
 962                        "br_netfilter: can't register to sysctl.\n");
 963                 nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
 964                 return -ENOMEM;
 965         }
 966 #endif
 967         printk(KERN_NOTICE "Bridge firewalling registered\n");
 968         return 0;
 969 }
 970
 971 void br_netfilter_fini(void)
 972 {
 973         nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
 974 #ifdef CONFIG_SYSCTL
 975         unregister_sysctl_table(brnf_sysctl_header);
 976 #endif
 977 }