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->tot_len = htons(nskb->len);
  99         ip_send_check(niph);
 100
 101         nft_reject_br_push_etherhdr(oldskb, nskb);
 102
 103         br_forward(br_port_get_rcu(dev), nskb, false, true);
 104 }
 105
 106 static void nft_reject_br_send_v4_unreach(struct net *net,
 107                                           struct sk_buff *oldskb,
 108                                           const struct net_device *dev,
 109                                           int hook, u8 code)
 110 {
 111         struct sk_buff *nskb;
 112         struct iphdr *niph;
 113         struct icmphdr *icmph;
 114         unsigned int len;
 115         __wsum csum;
 116         u8 proto;
 117
 118         if (!nft_bridge_iphdr_validate(oldskb))
 119                 return;
 120
 121         /* IP header checks: fragment. */
 122         if (ip_hdr(oldskb)->frag_off & htons(IP_OFFSET))
 123                 return;
 124
 125         /* RFC says return as much as we can without exceeding 576 bytes. */
 126         len = min_t(unsigned int, 536, oldskb->len);
 127
 128         if (!pskb_may_pull(oldskb, len))
 129                 return;
 130
 131         if (pskb_trim_rcsum(oldskb, ntohs(ip_hdr(oldskb)->tot_len)))
 132                 return;
 133
 134         if (ip_hdr(oldskb)->protocol == IPPROTO_TCP ||
 135             ip_hdr(oldskb)->protocol == IPPROTO_UDP)
 136                 proto = ip_hdr(oldskb)->protocol;
 137         else
 138                 proto = 0;
 139
 140         if (!skb_csum_unnecessary(oldskb) &&
 141             nf_ip_checksum(oldskb, hook, ip_hdrlen(oldskb), proto))
 142                 return;
 143
 144         nskb = alloc_skb(sizeof(struct iphdr) + sizeof(struct icmphdr) +
 145                          LL_MAX_HEADER + len, GFP_ATOMIC);
 146         if (!nskb)
 147                 return;
 148
 149         skb_reserve(nskb, LL_MAX_HEADER);
 150         niph = nf_reject_iphdr_put(nskb, oldskb, IPPROTO_ICMP,
 151                                    net->ipv4.sysctl_ip_default_ttl);
 152
 153         skb_reset_transport_header(nskb);
 154         icmph = skb_put_zero(nskb, sizeof(struct icmphdr));
 155         icmph->type     = ICMP_DEST_UNREACH;
 156         icmph->code     = code;
 157
 158         skb_put_data(nskb, skb_network_header(oldskb), len);
 159
 160         csum = csum_partial((void *)icmph, len + sizeof(struct icmphdr), 0);
 161         icmph->checksum = csum_fold(csum);
 162
 163         niph->tot_len   = htons(nskb->len);
 164         ip_send_check(niph);
 165
 166         nft_reject_br_push_etherhdr(oldskb, nskb);
 167
 168         br_forward(br_port_get_rcu(dev), nskb, false, true);
 169 }
 170
 171 static int nft_bridge_ip6hdr_validate(struct sk_buff *skb)
 172 {
 173         struct ipv6hdr *hdr;
 174         u32 pkt_len;
 175
 176         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 177                 return 0;
 178
 179         hdr = ipv6_hdr(skb);
 180         if (hdr->version != 6)
 181                 return 0;
 182
 183         pkt_len = ntohs(hdr->payload_len);
 184         if (pkt_len + sizeof(struct ipv6hdr) > skb->len)
 185                 return 0;
 186
 187         return 1;
 188 }
 189
 190 static void nft_reject_br_send_v6_tcp_reset(struct net *net,
 191                                             struct sk_buff *oldskb,
 192                                             const struct net_device *dev,
 193                                             int hook)
 194 {
 195         struct sk_buff *nskb;
 196         const struct tcphdr *oth;
 197         struct tcphdr _oth;
 198         unsigned int otcplen;
 199         struct ipv6hdr *nip6h;
 200
 201         if (!nft_bridge_ip6hdr_validate(oldskb))
 202                 return;
 203
 204         oth = nf_reject_ip6_tcphdr_get(oldskb, &_oth, &otcplen, hook);
 205         if (!oth)
 206                 return;
 207
 208         nskb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(struct tcphdr) +
 209                          LL_MAX_HEADER, GFP_ATOMIC);
 210         if (!nskb)
 211                 return;
 212
 213         skb_reserve(nskb, LL_MAX_HEADER);
 214         nip6h = nf_reject_ip6hdr_put(nskb, oldskb, IPPROTO_TCP,
 215                                      net->ipv6.devconf_all->hop_limit);
 216         nf_reject_ip6_tcphdr_put(nskb, oldskb, oth, otcplen);
 217         nip6h->payload_len = htons(nskb->len - sizeof(struct ipv6hdr));
 218
 219         nft_reject_br_push_etherhdr(oldskb, nskb);
 220
 221         br_forward(br_port_get_rcu(dev), nskb, false, true);
 222 }
 223
 224 static bool reject6_br_csum_ok(struct sk_buff *skb, int hook)
 225 {
 226         const struct ipv6hdr *ip6h = ipv6_hdr(skb);
 227         int thoff;
 228         __be16 fo;
 229         u8 proto = ip6h->nexthdr;
 230
 231         if (skb_csum_unnecessary(skb))
 232                 return true;
 233
 234         if (ip6h->payload_len &&
 235             pskb_trim_rcsum(skb, ntohs(ip6h->payload_len) + sizeof(*ip6h)))
 236                 return false;
 237
 238         ip6h = ipv6_hdr(skb);
 239         thoff = ipv6_skip_exthdr(skb, ((u8*)(ip6h+1) - skb->data), &proto, &fo);
 240         if (thoff < 0 || thoff >= skb->len || (fo & htons(~0x7)) != 0)
 241                 return false;
 242
 243         return nf_ip6_checksum(skb, hook, thoff, proto) == 0;
 244 }
 245
 246 static void nft_reject_br_send_v6_unreach(struct net *net,
 247                                           struct sk_buff *oldskb,
 248                                           const struct net_device *dev,
 249                                           int hook, u8 code)
 250 {
 251         struct sk_buff *nskb;
 252         struct ipv6hdr *nip6h;
 253         struct icmp6hdr *icmp6h;
 254         unsigned int len;
 255
 256         if (!nft_bridge_ip6hdr_validate(oldskb))
 257                 return;
 258
 259         /* Include "As much of invoking packet as possible without the ICMPv6
 260          * packet exceeding the minimum IPv6 MTU" in the ICMP payload.
 261          */
 262         len = min_t(unsigned int, 1220, oldskb->len);
 263
 264         if (!pskb_may_pull(oldskb, len))
 265                 return;
 266
 267         if (!reject6_br_csum_ok(oldskb, hook))
 268                 return;
 269
 270         nskb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(struct icmp6hdr) +
 271                          LL_MAX_HEADER + len, GFP_ATOMIC);
 272         if (!nskb)
 273                 return;
 274
 275         skb_reserve(nskb, LL_MAX_HEADER);
 276         nip6h = nf_reject_ip6hdr_put(nskb, oldskb, IPPROTO_ICMPV6,
 277                                      net->ipv6.devconf_all->hop_limit);
 278
 279         skb_reset_transport_header(nskb);
 280         icmp6h = skb_put_zero(nskb, sizeof(struct icmp6hdr));
 281         icmp6h->icmp6_type = ICMPV6_DEST_UNREACH;
 282         icmp6h->icmp6_code = code;
 283
 284         skb_put_data(nskb, skb_network_header(oldskb), len);
 285         nip6h->payload_len = htons(nskb->len - sizeof(struct ipv6hdr));
 286
 287         icmp6h->icmp6_cksum =
 288                 csum_ipv6_magic(&nip6h->saddr, &nip6h->daddr,
 289                                 nskb->len - sizeof(struct ipv6hdr),
 290                                 IPPROTO_ICMPV6,
 291                                 csum_partial(icmp6h,
 292                                              nskb->len - sizeof(struct ipv6hdr),
 293                                              0));
 294
 295         nft_reject_br_push_etherhdr(oldskb, nskb);
 296
 297         br_forward(br_port_get_rcu(dev), nskb, false, true);
 298 }
 299
 300 static void nft_reject_bridge_eval(const struct nft_expr *expr,
 301                                    struct nft_regs *regs,
 302                                    const struct nft_pktinfo *pkt)
 303 {
 304         struct nft_reject *priv = nft_expr_priv(expr);
 305         const unsigned char *dest = eth_hdr(pkt->skb)->h_dest;
 306
 307         if (is_broadcast_ether_addr(dest) ||
 308             is_multicast_ether_addr(dest))
 309                 goto out;
 310
 311         switch (eth_hdr(pkt->skb)->h_proto) {
 312         case htons(ETH_P_IP):
 313                 switch (priv->type) {
 314                 case NFT_REJECT_ICMP_UNREACH:
 315                         nft_reject_br_send_v4_unreach(nft_net(pkt), pkt->skb,
 316                                                       nft_in(pkt),
 317                                                       nft_hook(pkt),
 318                                                       priv->icmp_code);
 319                         break;
 320                 case NFT_REJECT_TCP_RST:
 321                         nft_reject_br_send_v4_tcp_reset(nft_net(pkt), pkt->skb,
 322                                                         nft_in(pkt),
 323                                                         nft_hook(pkt));
 324                         break;
 325                 case NFT_REJECT_ICMPX_UNREACH:
 326                         nft_reject_br_send_v4_unreach(nft_net(pkt), pkt->skb,
 327                                                       nft_in(pkt),
 328                                                       nft_hook(pkt),
 329                                                       nft_reject_icmp_code(priv->icmp_code));
 330                         break;
 331                 }
 332                 break;
 333         case htons(ETH_P_IPV6):
 334                 switch (priv->type) {
 335                 case NFT_REJECT_ICMP_UNREACH:
 336                         nft_reject_br_send_v6_unreach(nft_net(pkt), pkt->skb,
 337                                                       nft_in(pkt),
 338                                                       nft_hook(pkt),
 339                                                       priv->icmp_code);
 340                         break;
 341                 case NFT_REJECT_TCP_RST:
 342                         nft_reject_br_send_v6_tcp_reset(nft_net(pkt), pkt->skb,
 343                                                         nft_in(pkt),
 344                                                         nft_hook(pkt));
 345                         break;
 346                 case NFT_REJECT_ICMPX_UNREACH:
 347                         nft_reject_br_send_v6_unreach(nft_net(pkt), pkt->skb,
 348                                                       nft_in(pkt),
 349                                                       nft_hook(pkt),
 350                                                       nft_reject_icmpv6_code(priv->icmp_code));
 351                         break;
 352                 }
 353                 break;
 354         default:
 355                 /* No explicit way to reject this protocol, drop it. */
 356                 break;
 357         }
 358 out:
 359         regs->verdict.code = NF_DROP;
 360 }
 361
 362 static int nft_reject_bridge_validate(const struct nft_ctx *ctx,
 363                                       const struct nft_expr *expr,
 364                                       const struct nft_data **data)
 365 {
 366         return nft_chain_validate_hooks(ctx->chain, (1 << NF_BR_PRE_ROUTING) |
 367                                                     (1 << NF_BR_LOCAL_IN));
 368 }
 369
 370 static int nft_reject_bridge_init(const struct nft_ctx *ctx,
 371                                   const struct nft_expr *expr,
 372                                   const struct nlattr * const tb[])
 373 {
 374         struct nft_reject *priv = nft_expr_priv(expr);
 375         int icmp_code;
 376
 377         if (tb[NFTA_REJECT_TYPE] == NULL)
 378                 return -EINVAL;
 379
 380         priv->type = ntohl(nla_get_be32(tb[NFTA_REJECT_TYPE]));
 381         switch (priv->type) {
 382         case NFT_REJECT_ICMP_UNREACH:
 383         case NFT_REJECT_ICMPX_UNREACH:
 384                 if (tb[NFTA_REJECT_ICMP_CODE] == NULL)
 385                         return -EINVAL;
 386
 387                 icmp_code = nla_get_u8(tb[NFTA_REJECT_ICMP_CODE]);
 388                 if (priv->type == NFT_REJECT_ICMPX_UNREACH &&
 389                     icmp_code > NFT_REJECT_ICMPX_MAX)
 390                         return -EINVAL;
 391
 392                 priv->icmp_code = icmp_code;
 393                 break;
 394         case NFT_REJECT_TCP_RST:
 395                 break;
 396         default:
 397                 return -EINVAL;
 398         }
 399         return 0;
 400 }
 401
 402 static int nft_reject_bridge_dump(struct sk_buff *skb,
 403                                   const struct nft_expr *expr)
 404 {
 405         const struct nft_reject *priv = nft_expr_priv(expr);
 406
 407         if (nla_put_be32(skb, NFTA_REJECT_TYPE, htonl(priv->type)))
 408                 goto nla_put_failure;
 409
 410         switch (priv->type) {
 411         case NFT_REJECT_ICMP_UNREACH:
 412         case NFT_REJECT_ICMPX_UNREACH:
 413                 if (nla_put_u8(skb, NFTA_REJECT_ICMP_CODE, priv->icmp_code))
 414                         goto nla_put_failure;
 415                 break;
 416         default:
 417                 break;
 418         }
 419
 420         return 0;
 421
 422 nla_put_failure:
 423         return -1;
 424 }
 425
 426 static struct nft_expr_type nft_reject_bridge_type;
 427 static const struct nft_expr_ops nft_reject_bridge_ops = {
 428         .type           = &nft_reject_bridge_type,
 429         .size           = NFT_EXPR_SIZE(sizeof(struct nft_reject)),
 430         .eval           = nft_reject_bridge_eval,
 431         .init           = nft_reject_bridge_init,
 432         .dump           = nft_reject_bridge_dump,
 433         .validate       = nft_reject_bridge_validate,
 434 };
 435
 436 static struct nft_expr_type nft_reject_bridge_type __read_mostly = {
 437         .family         = NFPROTO_BRIDGE,
 438         .name           = "reject",
 439         .ops            = &nft_reject_bridge_ops,
 440         .policy         = nft_reject_policy,
 441         .maxattr        = NFTA_REJECT_MAX,
 442         .owner          = THIS_MODULE,
 443 };
 444
 445 static int __init nft_reject_bridge_module_init(void)
 446 {
 447         return nft_register_expr(&nft_reject_bridge_type);
 448 }
 449
 450 static void __exit nft_reject_bridge_module_exit(void)
 451 {
 452         nft_unregister_expr(&nft_reject_bridge_type);
 453 }
 454
 455 module_init(nft_reject_bridge_module_init);
 456 module_exit(nft_reject_bridge_module_exit);
 457
 458 MODULE_LICENSE("GPL");
 459 MODULE_AUTHOR("Pablo Neira Ayuso <pablo@netfilter.org>");
 460 MODULE_ALIAS_NFT_AF_EXPR(AF_BRIDGE, "reject");