net/bridge/netfilter/nft_reject_bridge.c

   1 /*
   2  * Copyright (c) 2014 Pablo Neira Ayuso <pablo@netfilter.org>
   3  *
   4  * This program is free software; you can redistribute it and/or modify
   5  * it under the terms of the GNU General Public License version 2 as
   6  * published by the Free Software Foundation.
   7  */
   8
   9 #include <linux/kernel.h>
  10 #include <linux/init.h>
  11 #include <linux/module.h>
  12 #include <linux/netlink.h>
  13 #include <linux/netfilter.h>
  14 #include <linux/netfilter/nf_tables.h>
  15 #include <net/netfilter/nf_tables.h>
  16 #include <net/netfilter/nft_reject.h>
  17 #include <net/netfilter/ipv4/nf_reject.h>
  18 #include <net/netfilter/ipv6/nf_reject.h>
  19 #include <linux/ip.h>
  20 #include <net/ip.h>
  21 #include <net/ip6_checksum.h>
  22 #include <linux/netfilter_bridge.h>
  23 #include <linux/netfilter_ipv6.h>
  24 #include "../br_private.h"
  25
  26 static void nft_reject_br_push_etherhdr(struct sk_buff *oldskb,
  27                                         struct sk_buff *nskb)
  28 {
  29         struct ethhdr *eth;
  30
  31         eth = skb_push(nskb, ETH_HLEN);
  32         skb_reset_mac_header(nskb);
  33         ether_addr_copy(eth->h_source, eth_hdr(oldskb)->h_dest);
  34         ether_addr_copy(eth->h_dest, eth_hdr(oldskb)->h_source);
  35         eth->h_proto = eth_hdr(oldskb)->h_proto;
  36         skb_pull(nskb, ETH_HLEN);
  37
  38         if (skb_vlan_tag_present(oldskb)) {
  39                 u16 vid = skb_vlan_tag_get(oldskb);
  40
  41                 __vlan_hwaccel_put_tag(nskb, oldskb->vlan_proto, vid);
  42         }
  43 }
  44
  45 static int nft_bridge_iphdr_validate(struct sk_buff *skb)
  46 {
  47         struct iphdr *iph;
  48         u32 len;
  49
  50         if (!pskb_may_pull(skb, sizeof(struct iphdr)))
  51                 return 0;
  52
  53         iph = ip_hdr(skb);
  54         if (iph->ihl < 5 || iph->version != 4)
  55                 return 0;
  56
  57         len = ntohs(iph->tot_len);
  58         if (skb->len < len)
  59                 return 0;
  60         else if (len < (iph->ihl*4))
  61                 return 0;
  62
  63         if (!pskb_may_pull(skb, iph->ihl*4))
  64                 return 0;
  65
  66         return 1;
  67 }
  68
  69 /* We cannot use oldskb->dev, it can be either bridge device (NF_BRIDGE INPUT)
  70  * or the bridge port (NF_BRIDGE PREROUTING).
  71  */
  72 static void nft_reject_br_send_v4_tcp_reset(struct net *net,
  73                                             struct sk_buff *oldskb,
  74                                             const struct net_device *dev,
  75                                             int hook)
  76 {
  77         struct sk_buff *nskb;
  78         struct iphdr *niph;
  79         const struct tcphdr *oth;
  80         struct tcphdr _oth;
  81
  82         if (!nft_bridge_iphdr_validate(oldskb))
  83                 return;
  84
  85         oth = nf_reject_ip_tcphdr_get(oldskb, &_oth, hook);
  86         if (!oth)
  87                 return;
  88
  89         nskb = alloc_skb(sizeof(struct iphdr) + sizeof(struct tcphdr) +
  90                          LL_MAX_HEADER, GFP_ATOMIC);
  91         if (!nskb)
  92                 return;
  93
  94         skb_reserve(nskb, LL_MAX_HEADER);
  95         niph = nf_reject_iphdr_put(nskb, oldskb, IPPROTO_TCP,
  96                                    net->ipv4.sysctl_ip_default_ttl);
  97         nf_reject_ip_tcphdr_put(nskb, oldskb, oth);
  98         niph->ttl       = net->ipv4.sysctl_ip_default_ttl;
  99         niph->tot_len   = htons(nskb->len);
 100         ip_send_check(niph);
 101
 102         nft_reject_br_push_etherhdr(oldskb, nskb);
 103
 104         br_forward(br_port_get_rcu(dev), nskb, false, true);
 105 }
 106
 107 static void nft_reject_br_send_v4_unreach(struct net *net,
 108                                           struct sk_buff *oldskb,
 109                                           const struct net_device *dev,
 110                                           int hook, u8 code)
 111 {
 112         struct sk_buff *nskb;
 113         struct iphdr *niph;
 114         struct icmphdr *icmph;
 115         unsigned int len;
 116         __wsum csum;
 117         u8 proto;
 118
 119         if (!nft_bridge_iphdr_validate(oldskb))
 120                 return;
 121
 122         /* IP header checks: fragment. */
 123         if (ip_hdr(oldskb)->frag_off & htons(IP_OFFSET))
 124                 return;
 125
 126         /* RFC says return as much as we can without exceeding 576 bytes. */
 127         len = min_t(unsigned int, 536, oldskb->len);
 128
 129         if (!pskb_may_pull(oldskb, len))
 130                 return;
 131
 132         if (pskb_trim_rcsum(oldskb, ntohs(ip_hdr(oldskb)->tot_len)))
 133                 return;
 134
 135         if (ip_hdr(oldskb)->protocol == IPPROTO_TCP ||
 136             ip_hdr(oldskb)->protocol == IPPROTO_UDP)
 137                 proto = ip_hdr(oldskb)->protocol;
 138         else
 139                 proto = 0;
 140
 141         if (!skb_csum_unnecessary(oldskb) &&
 142             nf_ip_checksum(oldskb, hook, ip_hdrlen(oldskb), proto))
 143                 return;
 144
 145         nskb = alloc_skb(sizeof(struct iphdr) + sizeof(struct icmphdr) +
 146                          LL_MAX_HEADER + len, GFP_ATOMIC);
 147         if (!nskb)
 148                 return;
 149
 150         skb_reserve(nskb, LL_MAX_HEADER);
 151         niph = nf_reject_iphdr_put(nskb, oldskb, IPPROTO_ICMP,
 152                                    net->ipv4.sysctl_ip_default_ttl);
 153
 154         skb_reset_transport_header(nskb);
 155         icmph = skb_put_zero(nskb, sizeof(struct icmphdr));
 156         icmph->type     = ICMP_DEST_UNREACH;
 157         icmph->code     = code;
 158
 159         skb_put_data(nskb, skb_network_header(oldskb), len);
 160
 161         csum = csum_partial((void *)icmph, len + sizeof(struct icmphdr), 0);
 162         icmph->checksum = csum_fold(csum);
 163
 164         niph->tot_len   = htons(nskb->len);
 165         ip_send_check(niph);
 166
 167         nft_reject_br_push_etherhdr(oldskb, nskb);
 168
 169         br_forward(br_port_get_rcu(dev), nskb, false, true);
 170 }
 171
 172 static int nft_bridge_ip6hdr_validate(struct sk_buff *skb)
 173 {
 174         struct ipv6hdr *hdr;
 175         u32 pkt_len;
 176
 177         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 178                 return 0;
 179
 180         hdr = ipv6_hdr(skb);
 181         if (hdr->version != 6)
 182                 return 0;
 183
 184         pkt_len = ntohs(hdr->payload_len);
 185         if (pkt_len + sizeof(struct ipv6hdr) > skb->len)
 186                 return 0;
 187
 188         return 1;
 189 }
 190
 191 static void nft_reject_br_send_v6_tcp_reset(struct net *net,
 192                                             struct sk_buff *oldskb,
 193                                             const struct net_device *dev,
 194                                             int hook)
 195 {
 196         struct sk_buff *nskb;
 197         const struct tcphdr *oth;
 198         struct tcphdr _oth;
 199         unsigned int otcplen;
 200         struct ipv6hdr *nip6h;
 201
 202         if (!nft_bridge_ip6hdr_validate(oldskb))
 203                 return;
 204
 205         oth = nf_reject_ip6_tcphdr_get(oldskb, &_oth, &otcplen, hook);
 206         if (!oth)
 207                 return;
 208
 209         nskb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(struct tcphdr) +
 210                          LL_MAX_HEADER, GFP_ATOMIC);
 211         if (!nskb)
 212                 return;
 213
 214         skb_reserve(nskb, LL_MAX_HEADER);
 215         nip6h = nf_reject_ip6hdr_put(nskb, oldskb, IPPROTO_TCP,
 216                                      net->ipv6.devconf_all->hop_limit);
 217         nf_reject_ip6_tcphdr_put(nskb, oldskb, oth, otcplen);
 218         nip6h->payload_len = htons(nskb->len - sizeof(struct ipv6hdr));
 219
 220         nft_reject_br_push_etherhdr(oldskb, nskb);
 221
 222         br_forward(br_port_get_rcu(dev), nskb, false, true);
 223 }
 224
 225 static bool reject6_br_csum_ok(struct sk_buff *skb, int hook)
 226 {
 227         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
 228         int thoff;
 229         __be16 fo;
 230         u8 proto = ip6h->nexthdr;
 231
 232         if (skb_csum_unnecessary(skb))
 233                 return true;
 234
 235         if (ip6h->payload_len &&
 236             pskb_trim_rcsum(skb, ntohs(ip6h->payload_len) + sizeof(*ip6h)))
 237                 return false;
 238
 239         ip6h = ipv6_hdr(skb);
 240         thoff = ipv6_skip_exthdr(skb, ((u8*)(ip6h+1) - skb->data), &proto, &fo);
 241         if (thoff < 0 || thoff >= skb->len || (fo & htons(~0x7)) != 0)
 242                 return false;
 243
 244         return nf_ip6_checksum(skb, hook, thoff, proto) == 0;
 245 }
 246
 247 static void nft_reject_br_send_v6_unreach(struct net *net,
 248                                           struct sk_buff *oldskb,
 249                                           const struct net_device *dev,
 250                                           int hook, u8 code)
 251 {
 252         struct sk_buff *nskb;
 253         struct ipv6hdr *nip6h;
 254         struct icmp6hdr *icmp6h;
 255         unsigned int len;
 256
 257         if (!nft_bridge_ip6hdr_validate(oldskb))
 258                 return;
 259
 260         /* Include "As much of invoking packet as possible without the ICMPv6
 261          * packet exceeding the minimum IPv6 MTU" in the ICMP payload.
 262          */
 263         len = min_t(unsigned int, 1220, oldskb->len);
 264
 265         if (!pskb_may_pull(oldskb, len))
 266                 return;
 267
 268         if (!reject6_br_csum_ok(oldskb, hook))
 269                 return;
 270
 271         nskb = alloc_skb(sizeof(struct iphdr) + sizeof(struct icmp6hdr) +
 272                          LL_MAX_HEADER + len, GFP_ATOMIC);
 273         if (!nskb)
 274                 return;
 275
 276         skb_reserve(nskb, LL_MAX_HEADER);
 277         nip6h = nf_reject_ip6hdr_put(nskb, oldskb, IPPROTO_ICMPV6,
 278                                      net->ipv6.devconf_all->hop_limit);
 279
 280         skb_reset_transport_header(nskb);
 281         icmp6h = skb_put_zero(nskb, sizeof(struct icmp6hdr));
 282         icmp6h->icmp6_type = ICMPV6_DEST_UNREACH;
 283         icmp6h->icmp6_code = code;
 284
 285         skb_put_data(nskb, skb_network_header(oldskb), len);
 286         nip6h->payload_len = htons(nskb->len - sizeof(struct ipv6hdr));
 287
 288         icmp6h->icmp6_cksum =
 289                 csum_ipv6_magic(&nip6h->saddr, &nip6h->daddr,
 290                                 nskb->len - sizeof(struct ipv6hdr),
 291                                 IPPROTO_ICMPV6,
 292                                 csum_partial(icmp6h,
 293                                              nskb->len - sizeof(struct ipv6hdr),
 294                                              0));
 295
 296         nft_reject_br_push_etherhdr(oldskb, nskb);
 297
 298         br_forward(br_port_get_rcu(dev), nskb, false, true);
 299 }
 300
 301 static void nft_reject_bridge_eval(const struct nft_expr *expr,
 302                                    struct nft_regs *regs,
 303                                    const struct nft_pktinfo *pkt)
 304 {
 305         struct nft_reject *priv = nft_expr_priv(expr);
 306         const unsigned char *dest = eth_hdr(pkt->skb)->h_dest;
 307
 308         if (is_broadcast_ether_addr(dest) ||
 309             is_multicast_ether_addr(dest))
 310                 goto out;
 311
 312         switch (eth_hdr(pkt->skb)->h_proto) {
 313         case htons(ETH_P_IP):
 314                 switch (priv->type) {
 315                 case NFT_REJECT_ICMP_UNREACH:
 316                         nft_reject_br_send_v4_unreach(nft_net(pkt), pkt->skb,
 317                                                       nft_in(pkt),
 318                                                       nft_hook(pkt),
 319                                                       priv->icmp_code);
 320                         break;
 321                 case NFT_REJECT_TCP_RST:
 322                         nft_reject_br_send_v4_tcp_reset(nft_net(pkt), pkt->skb,
 323                                                         nft_in(pkt),
 324                                                         nft_hook(pkt));
 325                         break;
 326                 case NFT_REJECT_ICMPX_UNREACH:
 327                         nft_reject_br_send_v4_unreach(nft_net(pkt), pkt->skb,
 328                                                       nft_in(pkt),
 329                                                       nft_hook(pkt),
 330                                                       nft_reject_icmp_code(priv->icmp_code));
 331                         break;
 332                 }
 333                 break;
 334         case htons(ETH_P_IPV6):
 335                 switch (priv->type) {
 336                 case NFT_REJECT_ICMP_UNREACH:
 337                         nft_reject_br_send_v6_unreach(nft_net(pkt), pkt->skb,
 338                                                       nft_in(pkt),
 339                                                       nft_hook(pkt),
 340                                                       priv->icmp_code);
 341                         break;
 342                 case NFT_REJECT_TCP_RST:
 343                         nft_reject_br_send_v6_tcp_reset(nft_net(pkt), pkt->skb,
 344                                                         nft_in(pkt),
 345                                                         nft_hook(pkt));
 346                         break;
 347                 case NFT_REJECT_ICMPX_UNREACH:
 348                         nft_reject_br_send_v6_unreach(nft_net(pkt), pkt->skb,
 349                                                       nft_in(pkt),
 350                                                       nft_hook(pkt),
 351                                                       nft_reject_icmpv6_code(priv->icmp_code));
 352                         break;
 353                 }
 354                 break;
 355         default:
 356                 /* No explicit way to reject this protocol, drop it. */
 357                 break;
 358         }
 359 out:
 360         regs->verdict.code = NF_DROP;
 361 }
 362
 363 static int nft_reject_bridge_validate(const struct nft_ctx *ctx,
 364                                       const struct nft_expr *expr,
 365                                       const struct nft_data **data)
 366 {
 367         return nft_chain_validate_hooks(ctx->chain, (1 << NF_BR_PRE_ROUTING) |
 368                                                     (1 << NF_BR_LOCAL_IN));
 369 }
 370
 371 static int nft_reject_bridge_init(const struct nft_ctx *ctx,
 372                                   const struct nft_expr *expr,
 373                                   const struct nlattr * const tb[])
 374 {
 375         struct nft_reject *priv = nft_expr_priv(expr);
 376         int icmp_code;
 377
 378         if (tb[NFTA_REJECT_TYPE] == NULL)
 379                 return -EINVAL;
 380
 381         priv->type = ntohl(nla_get_be32(tb[NFTA_REJECT_TYPE]));
 382         switch (priv->type) {
 383         case NFT_REJECT_ICMP_UNREACH:
 384         case NFT_REJECT_ICMPX_UNREACH:
 385                 if (tb[NFTA_REJECT_ICMP_CODE] == NULL)
 386                         return -EINVAL;
 387
 388                 icmp_code = nla_get_u8(tb[NFTA_REJECT_ICMP_CODE]);
 389                 if (priv->type == NFT_REJECT_ICMPX_UNREACH &&
 390                     icmp_code > NFT_REJECT_ICMPX_MAX)
 391                         return -EINVAL;
 392
 393                 priv->icmp_code = icmp_code;
 394                 break;
 395         case NFT_REJECT_TCP_RST:
 396                 break;
 397         default:
 398                 return -EINVAL;
 399         }
 400         return 0;
 401 }
 402
 403 static int nft_reject_bridge_dump(struct sk_buff *skb,
 404                                   const struct nft_expr *expr)
 405 {
 406         const struct nft_reject *priv = nft_expr_priv(expr);
 407
 408         if (nla_put_be32(skb, NFTA_REJECT_TYPE, htonl(priv->type)))
 409                 goto nla_put_failure;
 410
 411         switch (priv->type) {
 412         case NFT_REJECT_ICMP_UNREACH:
 413         case NFT_REJECT_ICMPX_UNREACH:
 414                 if (nla_put_u8(skb, NFTA_REJECT_ICMP_CODE, priv->icmp_code))
 415                         goto nla_put_failure;
 416                 break;
 417         default:
 418                 break;
 419         }
 420
 421         return 0;
 422
 423 nla_put_failure:
 424         return -1;
 425 }
 426
 427 static struct nft_expr_type nft_reject_bridge_type;
 428 static const struct nft_expr_ops nft_reject_bridge_ops = {
 429         .type           = &nft_reject_bridge_type,
 430         .size           = NFT_EXPR_SIZE(sizeof(struct nft_reject)),
 431         .eval           = nft_reject_bridge_eval,
 432         .init           = nft_reject_bridge_init,
 433         .dump           = nft_reject_bridge_dump,
 434         .validate       = nft_reject_bridge_validate,
 435 };
 436
 437 static struct nft_expr_type nft_reject_bridge_type __read_mostly = {
 438         .family         = NFPROTO_BRIDGE,
 439         .name           = "reject",
 440         .ops            = &nft_reject_bridge_ops,
 441         .policy         = nft_reject_policy,
 442         .maxattr        = NFTA_REJECT_MAX,
 443         .owner          = THIS_MODULE,
 444 };
 445
 446 static int __init nft_reject_bridge_module_init(void)
 447 {
 448         return nft_register_expr(&nft_reject_bridge_type);
 449 }
 450
 451 static void __exit nft_reject_bridge_module_exit(void)
 452 {
 453         nft_unregister_expr(&nft_reject_bridge_type);
 454 }
 455
 456 module_init(nft_reject_bridge_module_init);
 457 module_exit(nft_reject_bridge_module_exit);
 458
 459 MODULE_LICENSE("GPL");
 460 MODULE_AUTHOR("Pablo Neira Ayuso <pablo@netfilter.org>");
 461 MODULE_ALIAS_NFT_AF_EXPR(AF_BRIDGE, "reject");