net/sched/act_nat.c

   1 /*
   2  * Stateless NAT actions
   3  *
   4  * Copyright (c) 2007 Herbert Xu <herbert@gondor.apana.org.au>
   5  *
   6  * This program is free software; you can redistribute it and/or modify it
   7  * under the terms of the GNU General Public License as published by the Free
   8  * Software Foundation; either version 2 of the License, or (at your option)
   9  * any later version.
  10  */
  11
  12 #include <linux/errno.h>
  13 #include <linux/init.h>
  14 #include <linux/kernel.h>
  15 #include <linux/module.h>
  16 #include <linux/netfilter.h>
  17 #include <linux/rtnetlink.h>
  18 #include <linux/skbuff.h>
  19 #include <linux/slab.h>
  20 #include <linux/spinlock.h>
  21 #include <linux/string.h>
  22 #include <linux/tc_act/tc_nat.h>
  23 #include <net/act_api.h>
  24 #include <net/icmp.h>
  25 #include <net/ip.h>
  26 #include <net/netlink.h>
  27 #include <net/tc_act/tc_nat.h>
  28 #include <net/tcp.h>
  29 #include <net/udp.h>
  30
  31
  32 #define NAT_TAB_MASK    15
  33
  34 static int nat_net_id;
  35
  36 static const struct nla_policy nat_policy[TCA_NAT_MAX + 1] = {
  37         [TCA_NAT_PARMS] = { .len = sizeof(struct tc_nat) },
  38 };
  39
  40 static int tcf_nat_init(struct net *net, struct nlattr *nla, struct nlattr *est,
  41                         struct tc_action *a, int ovr, int bind)
  42 {
  43         struct tc_action_net *tn = net_generic(net, nat_net_id);
  44         struct nlattr *tb[TCA_NAT_MAX + 1];
  45         struct tc_nat *parm;
  46         int ret = 0, err;
  47         struct tcf_nat *p;
  48
  49         if (nla == NULL)
  50                 return -EINVAL;
  51
  52         err = nla_parse_nested(tb, TCA_NAT_MAX, nla, nat_policy);
  53         if (err < 0)
  54                 return err;
  55
  56         if (tb[TCA_NAT_PARMS] == NULL)
  57                 return -EINVAL;
  58         parm = nla_data(tb[TCA_NAT_PARMS]);
  59
  60         if (!tcf_hash_check(tn, parm->index, a, bind)) {
  61                 ret = tcf_hash_create(tn, parm->index, est, a,
  62                                       sizeof(*p), bind, false);
  63                 if (ret)
  64                         return ret;
  65                 ret = ACT_P_CREATED;
  66         } else {
  67                 if (bind)
  68                         return 0;
  69                 tcf_hash_release(a, bind);
  70                 if (!ovr)
  71                         return -EEXIST;
  72         }
  73         p = to_tcf_nat(a);
  74
  75         spin_lock_bh(&p->tcf_lock);
  76         p->old_addr = parm->old_addr;
  77         p->new_addr = parm->new_addr;
  78         p->mask = parm->mask;
  79         p->flags = parm->flags;
  80
  81         p->tcf_action = parm->action;
  82         spin_unlock_bh(&p->tcf_lock);
  83
  84         if (ret == ACT_P_CREATED)
  85                 tcf_hash_insert(tn, a);
  86
  87         return ret;
  88 }
  89
  90 static int tcf_nat(struct sk_buff *skb, const struct tc_action *a,
  91                    struct tcf_result *res)
  92 {
  93         struct tcf_nat *p = a->priv;
  94         struct iphdr *iph;
  95         __be32 old_addr;
  96         __be32 new_addr;
  97         __be32 mask;
  98         __be32 addr;
  99         int egress;
 100         int action;
 101         int ihl;
 102         int noff;
 103
 104         spin_lock(&p->tcf_lock);
 105
 106         p->tcf_tm.lastuse = jiffies;
 107         old_addr = p->old_addr;
 108         new_addr = p->new_addr;
 109         mask = p->mask;
 110         egress = p->flags & TCA_NAT_FLAG_EGRESS;
 111         action = p->tcf_action;
 112
 113         bstats_update(&p->tcf_bstats, skb);
 114
 115         spin_unlock(&p->tcf_lock);
 116
 117         if (unlikely(action == TC_ACT_SHOT))
 118                 goto drop;
 119
 120         noff = skb_network_offset(skb);
 121         if (!pskb_may_pull(skb, sizeof(*iph) + noff))
 122                 goto drop;
 123
 124         iph = ip_hdr(skb);
 125
 126         if (egress)
 127                 addr = iph->saddr;
 128         else
 129                 addr = iph->daddr;
 130
 131         if (!((old_addr ^ addr) & mask)) {
 132                 if (skb_try_make_writable(skb, sizeof(*iph) + noff))
 133                         goto drop;
 134
 135                 new_addr &= mask;
 136                 new_addr |= addr & ~mask;
 137
 138                 /* Rewrite IP header */
 139                 iph = ip_hdr(skb);
 140                 if (egress)
 141                         iph->saddr = new_addr;
 142                 else
 143                         iph->daddr = new_addr;
 144
 145                 csum_replace4(&iph->check, addr, new_addr);
 146         } else if ((iph->frag_off & htons(IP_OFFSET)) ||
 147                    iph->protocol != IPPROTO_ICMP) {
 148                 goto out;
 149         }
 150
 151         ihl = iph->ihl * 4;
 152
 153         /* It would be nice to share code with stateful NAT. */
 154         switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
 155         case IPPROTO_TCP:
 156         {
 157                 struct tcphdr *tcph;
 158
 159                 if (!pskb_may_pull(skb, ihl + sizeof(*tcph) + noff) ||
 160                     skb_try_make_writable(skb, ihl + sizeof(*tcph) + noff))
 161                         goto drop;
 162
 163                 tcph = (void *)(skb_network_header(skb) + ihl);
 164                 inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr,
 165                                          true);
 166                 break;
 167         }
 168         case IPPROTO_UDP:
 169         {
 170                 struct udphdr *udph;
 171
 172                 if (!pskb_may_pull(skb, ihl + sizeof(*udph) + noff) ||
 173                     skb_try_make_writable(skb, ihl + sizeof(*udph) + noff))
 174                         goto drop;
 175
 176                 udph = (void *)(skb_network_header(skb) + ihl);
 177                 if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
 178                         inet_proto_csum_replace4(&udph->check, skb, addr,
 179                                                  new_addr, true);
 180                         if (!udph->check)
 181                                 udph->check = CSUM_MANGLED_0;
 182                 }
 183                 break;
 184         }
 185         case IPPROTO_ICMP:
 186         {
 187                 struct icmphdr *icmph;
 188
 189                 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + noff))
 190                         goto drop;
 191
 192                 icmph = (void *)(skb_network_header(skb) + ihl);
 193
 194                 if ((icmph->type != ICMP_DEST_UNREACH) &&
 195                     (icmph->type != ICMP_TIME_EXCEEDED) &&
 196                     (icmph->type != ICMP_PARAMETERPROB))
 197                         break;
 198
 199                 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph) +
 200                                         noff))
 201                         goto drop;
 202
 203                 icmph = (void *)(skb_network_header(skb) + ihl);
 204                 iph = (void *)(icmph + 1);
 205                 if (egress)
 206                         addr = iph->daddr;
 207                 else
 208                         addr = iph->saddr;
 209
 210                 if ((old_addr ^ addr) & mask)
 211                         break;
 212
 213                 if (skb_try_make_writable(skb, ihl + sizeof(*icmph) +
 214                                           sizeof(*iph) + noff))
 215                         goto drop;
 216
 217                 icmph = (void *)(skb_network_header(skb) + ihl);
 218                 iph = (void *)(icmph + 1);
 219
 220                 new_addr &= mask;
 221                 new_addr |= addr & ~mask;
 222
 223                 /* XXX Fix up the inner checksums. */
 224                 if (egress)
 225                         iph->daddr = new_addr;
 226                 else
 227                         iph->saddr = new_addr;
 228
 229                 inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
 230                                          false);
 231                 break;
 232         }
 233         default:
 234                 break;
 235         }
 236
 237 out:
 238         return action;
 239
 240 drop:
 241         spin_lock(&p->tcf_lock);
 242         p->tcf_qstats.drops++;
 243         spin_unlock(&p->tcf_lock);
 244         return TC_ACT_SHOT;
 245 }
 246
 247 static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
 248                         int bind, int ref)
 249 {
 250         unsigned char *b = skb_tail_pointer(skb);
 251         struct tcf_nat *p = a->priv;
 252         struct tc_nat opt = {
 253                 .old_addr = p->old_addr,
 254                 .new_addr = p->new_addr,
 255                 .mask     = p->mask,
 256                 .flags    = p->flags,
 257
 258                 .index    = p->tcf_index,
 259                 .action   = p->tcf_action,
 260                 .refcnt   = p->tcf_refcnt - ref,
 261                 .bindcnt  = p->tcf_bindcnt - bind,
 262         };
 263         struct tcf_t t;
 264
 265         if (nla_put(skb, TCA_NAT_PARMS, sizeof(opt), &opt))
 266                 goto nla_put_failure;
 267         t.install = jiffies_to_clock_t(jiffies - p->tcf_tm.install);
 268         t.lastuse = jiffies_to_clock_t(jiffies - p->tcf_tm.lastuse);
 269         t.expires = jiffies_to_clock_t(p->tcf_tm.expires);
 270         if (nla_put(skb, TCA_NAT_TM, sizeof(t), &t))
 271                 goto nla_put_failure;
 272
 273         return skb->len;
 274
 275 nla_put_failure:
 276         nlmsg_trim(skb, b);
 277         return -1;
 278 }
 279
 280 static int tcf_nat_walker(struct net *net, struct sk_buff *skb,
 281                           struct netlink_callback *cb, int type,
 282                           struct tc_action *a)
 283 {
 284         struct tc_action_net *tn = net_generic(net, nat_net_id);
 285
 286         return tcf_generic_walker(tn, skb, cb, type, a);
 287 }
 288
 289 static int tcf_nat_search(struct net *net, struct tc_action *a, u32 index)
 290 {
 291         struct tc_action_net *tn = net_generic(net, nat_net_id);
 292
 293         return tcf_hash_search(tn, a, index);
 294 }
 295
 296 static struct tc_action_ops act_nat_ops = {
 297         .kind           =       "nat",
 298         .type           =       TCA_ACT_NAT,
 299         .owner          =       THIS_MODULE,
 300         .act            =       tcf_nat,
 301         .dump           =       tcf_nat_dump,
 302         .init           =       tcf_nat_init,
 303         .walk           =       tcf_nat_walker,
 304         .lookup         =       tcf_nat_search,
 305 };
 306
 307 static __net_init int nat_init_net(struct net *net)
 308 {
 309         struct tc_action_net *tn = net_generic(net, nat_net_id);
 310
 311         return tc_action_net_init(tn, &act_nat_ops, NAT_TAB_MASK);
 312 }
 313
 314 static void __net_exit nat_exit_net(struct net *net)
 315 {
 316         struct tc_action_net *tn = net_generic(net, nat_net_id);
 317
 318         tc_action_net_exit(tn);
 319 }
 320
 321 static struct pernet_operations nat_net_ops = {
 322         .init = nat_init_net,
 323         .exit = nat_exit_net,
 324         .id   = &nat_net_id,
 325         .size = sizeof(struct tc_action_net),
 326 };
 327
 328 MODULE_DESCRIPTION("Stateless NAT actions");
 329 MODULE_LICENSE("GPL");
 330
 331 static int __init nat_init_module(void)
 332 {
 333         return tcf_register_action(&act_nat_ops, &nat_net_ops);
 334 }
 335
 336 static void __exit nat_cleanup_module(void)
 337 {
 338         tcf_unregister_action(&act_nat_ops, &nat_net_ops);
 339 }
 340
 341 module_init(nat_init_module);
 342 module_exit(nat_cleanup_module);