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 static unsigned int nat_net_id;
  33 static struct tc_action_ops act_nat_ops;
  34
  35 static const struct nla_policy nat_policy[TCA_NAT_MAX + 1] = {
  36         [TCA_NAT_PARMS] = { .len = sizeof(struct tc_nat) },
  37 };
  38
  39 static int tcf_nat_init(struct net *net, struct nlattr *nla, struct nlattr *est,
  40                         struct tc_action **a, int ovr, int bind)
  41 {
  42         struct tc_action_net *tn = net_generic(net, nat_net_id);
  43         struct nlattr *tb[TCA_NAT_MAX + 1];
  44         struct tc_nat *parm;
  45         int ret = 0, err;
  46         struct tcf_nat *p;
  47
  48         if (nla == NULL)
  49                 return -EINVAL;
  50
  51         err = nla_parse_nested(tb, TCA_NAT_MAX, nla, nat_policy, NULL);
  52         if (err < 0)
  53                 return err;
  54
  55         if (tb[TCA_NAT_PARMS] == NULL)
  56                 return -EINVAL;
  57         parm = nla_data(tb[TCA_NAT_PARMS]);
  58
  59         if (!tcf_idr_check(tn, parm->index, a, bind)) {
  60                 ret = tcf_idr_create(tn, parm->index, est, a,
  61                                      &act_nat_ops, bind, false);
  62                 if (ret)
  63                         return ret;
  64                 ret = ACT_P_CREATED;
  65         } else {
  66                 if (bind)
  67                         return 0;
  68                 tcf_idr_release(*a, bind);
  69                 if (!ovr)
  70                         return -EEXIST;
  71         }
  72         p = to_tcf_nat(*a);
  73
  74         spin_lock_bh(&p->tcf_lock);
  75         p->old_addr = parm->old_addr;
  76         p->new_addr = parm->new_addr;
  77         p->mask = parm->mask;
  78         p->flags = parm->flags;
  79
  80         p->tcf_action = parm->action;
  81         spin_unlock_bh(&p->tcf_lock);
  82
  83         if (ret == ACT_P_CREATED)
  84                 tcf_idr_insert(tn, *a);
  85
  86         return ret;
  87 }
  88
  89 static int tcf_nat(struct sk_buff *skb, const struct tc_action *a,
  90                    struct tcf_result *res)
  91 {
  92         struct tcf_nat *p = to_tcf_nat(a);
  93         struct iphdr *iph;
  94         __be32 old_addr;
  95         __be32 new_addr;
  96         __be32 mask;
  97         __be32 addr;
  98         int egress;
  99         int action;
 100         int ihl;
 101         int noff;
 102
 103         spin_lock(&p->tcf_lock);
 104
 105         tcf_lastuse_update(&p->tcf_tm);
 106         old_addr = p->old_addr;
 107         new_addr = p->new_addr;
 108         mask = p->mask;
 109         egress = p->flags & TCA_NAT_FLAG_EGRESS;
 110         action = p->tcf_action;
 111
 112         bstats_update(&p->tcf_bstats, skb);
 113
 114         spin_unlock(&p->tcf_lock);
 115
 116         if (unlikely(action == TC_ACT_SHOT))
 117                 goto drop;
 118
 119         noff = skb_network_offset(skb);
 120         if (!pskb_may_pull(skb, sizeof(*iph) + noff))
 121                 goto drop;
 122
 123         iph = ip_hdr(skb);
 124
 125         if (egress)
 126                 addr = iph->saddr;
 127         else
 128                 addr = iph->daddr;
 129
 130         if (!((old_addr ^ addr) & mask)) {
 131                 if (skb_try_make_writable(skb, sizeof(*iph) + noff))
 132                         goto drop;
 133
 134                 new_addr &= mask;
 135                 new_addr |= addr & ~mask;
 136
 137                 /* Rewrite IP header */
 138                 iph = ip_hdr(skb);
 139                 if (egress)
 140                         iph->saddr = new_addr;
 141                 else
 142                         iph->daddr = new_addr;
 143
 144                 csum_replace4(&iph->check, addr, new_addr);
 145         } else if ((iph->frag_off & htons(IP_OFFSET)) ||
 146                    iph->protocol != IPPROTO_ICMP) {
 147                 goto out;
 148         }
 149
 150         ihl = iph->ihl * 4;
 151
 152         /* It would be nice to share code with stateful NAT. */
 153         switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
 154         case IPPROTO_TCP:
 155         {
 156                 struct tcphdr *tcph;
 157
 158                 if (!pskb_may_pull(skb, ihl + sizeof(*tcph) + noff) ||
 159                     skb_try_make_writable(skb, ihl + sizeof(*tcph) + noff))
 160                         goto drop;
 161
 162                 tcph = (void *)(skb_network_header(skb) + ihl);
 163                 inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr,
 164                                          true);
 165                 break;
 166         }
 167         case IPPROTO_UDP:
 168         {
 169                 struct udphdr *udph;
 170
 171                 if (!pskb_may_pull(skb, ihl + sizeof(*udph) + noff) ||
 172                     skb_try_make_writable(skb, ihl + sizeof(*udph) + noff))
 173                         goto drop;
 174
 175                 udph = (void *)(skb_network_header(skb) + ihl);
 176                 if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
 177                         inet_proto_csum_replace4(&udph->check, skb, addr,
 178                                                  new_addr, true);
 179                         if (!udph->check)
 180                                 udph->check = CSUM_MANGLED_0;
 181                 }
 182                 break;
 183         }
 184         case IPPROTO_ICMP:
 185         {
 186                 struct icmphdr *icmph;
 187
 188                 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + noff))
 189                         goto drop;
 190
 191                 icmph = (void *)(skb_network_header(skb) + ihl);
 192
 193                 if ((icmph->type != ICMP_DEST_UNREACH) &&
 194                     (icmph->type != ICMP_TIME_EXCEEDED) &&
 195                     (icmph->type != ICMP_PARAMETERPROB))
 196                         break;
 197
 198                 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph) +
 199                                         noff))
 200                         goto drop;
 201
 202                 icmph = (void *)(skb_network_header(skb) + ihl);
 203                 iph = (void *)(icmph + 1);
 204                 if (egress)
 205                         addr = iph->daddr;
 206                 else
 207                         addr = iph->saddr;
 208
 209                 if ((old_addr ^ addr) & mask)
 210                         break;
 211
 212                 if (skb_try_make_writable(skb, ihl + sizeof(*icmph) +
 213                                           sizeof(*iph) + noff))
 214                         goto drop;
 215
 216                 icmph = (void *)(skb_network_header(skb) + ihl);
 217                 iph = (void *)(icmph + 1);
 218
 219                 new_addr &= mask;
 220                 new_addr |= addr & ~mask;
 221
 222                 /* XXX Fix up the inner checksums. */
 223                 if (egress)
 224                         iph->daddr = new_addr;
 225                 else
 226                         iph->saddr = new_addr;
 227
 228                 inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
 229                                          false);
 230                 break;
 231         }
 232         default:
 233                 break;
 234         }
 235
 236 out:
 237         return action;
 238
 239 drop:
 240         spin_lock(&p->tcf_lock);
 241         p->tcf_qstats.drops++;
 242         spin_unlock(&p->tcf_lock);
 243         return TC_ACT_SHOT;
 244 }
 245
 246 static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
 247                         int bind, int ref)
 248 {
 249         unsigned char *b = skb_tail_pointer(skb);
 250         struct tcf_nat *p = to_tcf_nat(a);
 251         struct tc_nat opt = {
 252                 .old_addr = p->old_addr,
 253                 .new_addr = p->new_addr,
 254                 .mask     = p->mask,
 255                 .flags    = p->flags,
 256
 257                 .index    = p->tcf_index,
 258                 .action   = p->tcf_action,
 259                 .refcnt   = p->tcf_refcnt - ref,
 260                 .bindcnt  = p->tcf_bindcnt - bind,
 261         };
 262         struct tcf_t t;
 263
 264         if (nla_put(skb, TCA_NAT_PARMS, sizeof(opt), &opt))
 265                 goto nla_put_failure;
 266
 267         tcf_tm_dump(&t, &p->tcf_tm);
 268         if (nla_put_64bit(skb, TCA_NAT_TM, sizeof(t), &t, TCA_NAT_PAD))
 269                 goto nla_put_failure;
 270
 271         return skb->len;
 272
 273 nla_put_failure:
 274         nlmsg_trim(skb, b);
 275         return -1;
 276 }
 277
 278 static int tcf_nat_walker(struct net *net, struct sk_buff *skb,
 279                           struct netlink_callback *cb, int type,
 280                           const struct tc_action_ops *ops)
 281 {
 282         struct tc_action_net *tn = net_generic(net, nat_net_id);
 283
 284         return tcf_generic_walker(tn, skb, cb, type, ops);
 285 }
 286
 287 static int tcf_nat_search(struct net *net, struct tc_action **a, u32 index)
 288 {
 289         struct tc_action_net *tn = net_generic(net, nat_net_id);
 290
 291         return tcf_idr_search(tn, a, index);
 292 }
 293
 294 static struct tc_action_ops act_nat_ops = {
 295         .kind           =       "nat",
 296         .type           =       TCA_ACT_NAT,
 297         .owner          =       THIS_MODULE,
 298         .act            =       tcf_nat,
 299         .dump           =       tcf_nat_dump,
 300         .init           =       tcf_nat_init,
 301         .walk           =       tcf_nat_walker,
 302         .lookup         =       tcf_nat_search,
 303         .size           =       sizeof(struct tcf_nat),
 304 };
 305
 306 static __net_init int nat_init_net(struct net *net)
 307 {
 308         struct tc_action_net *tn = net_generic(net, nat_net_id);
 309
 310         return tc_action_net_init(tn, &act_nat_ops);
 311 }
 312
 313 static void __net_exit nat_exit_net(struct list_head *net_list)
 314 {
 315         tc_action_net_exit(net_list, nat_net_id);
 316 }
 317
 318 static struct pernet_operations nat_net_ops = {
 319         .init = nat_init_net,
 320         .exit_batch = nat_exit_net,
 321         .id   = &nat_net_id,
 322         .size = sizeof(struct tc_action_net),
 323 };
 324
 325 MODULE_DESCRIPTION("Stateless NAT actions");
 326 MODULE_LICENSE("GPL");
 327
 328 static int __init nat_init_module(void)
 329 {
 330         return tcf_register_action(&act_nat_ops, &nat_net_ops);
 331 }
 332
 333 static void __exit nat_cleanup_module(void)
 334 {
 335         tcf_unregister_action(&act_nat_ops, &nat_net_ops);
 336 }
 337
 338 module_init(nat_init_module);
 339 module_exit(nat_cleanup_module);