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