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                         bool rtnl_held, struct netlink_ext_ack *extack)
  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         u32 index;
  49
  50         if (nla == NULL)
  51                 return -EINVAL;
  52
  53         err = nla_parse_nested(tb, TCA_NAT_MAX, nla, nat_policy, NULL);
  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         index = parm->index;
  61         err = tcf_idr_check_alloc(tn, &index, a, bind);
  62         if (!err) {
  63                 ret = tcf_idr_create(tn, index, est, a,
  64                                      &act_nat_ops, bind, false);
  65                 if (ret) {
  66                         tcf_idr_cleanup(tn, index);
  67                         return ret;
  68                 }
  69                 ret = ACT_P_CREATED;
  70         } else if (err > 0) {
  71                 if (bind)
  72                         return 0;
  73                 if (!ovr) {
  74                         tcf_idr_release(*a, bind);
  75                         return -EEXIST;
  76                 }
  77         } else {
  78                 return err;
  79         }
  80         p = to_tcf_nat(*a);
  81
  82         spin_lock_bh(&p->tcf_lock);
  83         p->old_addr = parm->old_addr;
  84         p->new_addr = parm->new_addr;
  85         p->mask = parm->mask;
  86         p->flags = parm->flags;
  87
  88         p->tcf_action = parm->action;
  89         spin_unlock_bh(&p->tcf_lock);
  90
  91         if (ret == ACT_P_CREATED)
  92                 tcf_idr_insert(tn, *a);
  93
  94         return ret;
  95 }
  96
  97 static int tcf_nat_act(struct sk_buff *skb, const struct tc_action *a,
  98                        struct tcf_result *res)
  99 {
 100         struct tcf_nat *p = to_tcf_nat(a);
 101         struct iphdr *iph;
 102         __be32 old_addr;
 103         __be32 new_addr;
 104         __be32 mask;
 105         __be32 addr;
 106         int egress;
 107         int action;
 108         int ihl;
 109         int noff;
 110
 111         spin_lock(&p->tcf_lock);
 112
 113         tcf_lastuse_update(&p->tcf_tm);
 114         old_addr = p->old_addr;
 115         new_addr = p->new_addr;
 116         mask = p->mask;
 117         egress = p->flags & TCA_NAT_FLAG_EGRESS;
 118         action = p->tcf_action;
 119
 120         bstats_update(&p->tcf_bstats, skb);
 121
 122         spin_unlock(&p->tcf_lock);
 123
 124         if (unlikely(action == TC_ACT_SHOT))
 125                 goto drop;
 126
 127         noff = skb_network_offset(skb);
 128         if (!pskb_may_pull(skb, sizeof(*iph) + noff))
 129                 goto drop;
 130
 131         iph = ip_hdr(skb);
 132
 133         if (egress)
 134                 addr = iph->saddr;
 135         else
 136                 addr = iph->daddr;
 137
 138         if (!((old_addr ^ addr) & mask)) {
 139                 if (skb_try_make_writable(skb, sizeof(*iph) + noff))
 140                         goto drop;
 141
 142                 new_addr &= mask;
 143                 new_addr |= addr & ~mask;
 144
 145                 /* Rewrite IP header */
 146                 iph = ip_hdr(skb);
 147                 if (egress)
 148                         iph->saddr = new_addr;
 149                 else
 150                         iph->daddr = new_addr;
 151
 152                 csum_replace4(&iph->check, addr, new_addr);
 153         } else if ((iph->frag_off & htons(IP_OFFSET)) ||
 154                    iph->protocol != IPPROTO_ICMP) {
 155                 goto out;
 156         }
 157
 158         ihl = iph->ihl * 4;
 159
 160         /* It would be nice to share code with stateful NAT. */
 161         switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
 162         case IPPROTO_TCP:
 163         {
 164                 struct tcphdr *tcph;
 165
 166                 if (!pskb_may_pull(skb, ihl + sizeof(*tcph) + noff) ||
 167                     skb_try_make_writable(skb, ihl + sizeof(*tcph) + noff))
 168                         goto drop;
 169
 170                 tcph = (void *)(skb_network_header(skb) + ihl);
 171                 inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr,
 172                                          true);
 173                 break;
 174         }
 175         case IPPROTO_UDP:
 176         {
 177                 struct udphdr *udph;
 178
 179                 if (!pskb_may_pull(skb, ihl + sizeof(*udph) + noff) ||
 180                     skb_try_make_writable(skb, ihl + sizeof(*udph) + noff))
 181                         goto drop;
 182
 183                 udph = (void *)(skb_network_header(skb) + ihl);
 184                 if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
 185                         inet_proto_csum_replace4(&udph->check, skb, addr,
 186                                                  new_addr, true);
 187                         if (!udph->check)
 188                                 udph->check = CSUM_MANGLED_0;
 189                 }
 190                 break;
 191         }
 192         case IPPROTO_ICMP:
 193         {
 194                 struct icmphdr *icmph;
 195
 196                 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + noff))
 197                         goto drop;
 198
 199                 icmph = (void *)(skb_network_header(skb) + ihl);
 200
 201                 if ((icmph->type != ICMP_DEST_UNREACH) &&
 202                     (icmph->type != ICMP_TIME_EXCEEDED) &&
 203                     (icmph->type != ICMP_PARAMETERPROB))
 204                         break;
 205
 206                 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph) +
 207                                         noff))
 208                         goto drop;
 209
 210                 icmph = (void *)(skb_network_header(skb) + ihl);
 211                 iph = (void *)(icmph + 1);
 212                 if (egress)
 213                         addr = iph->daddr;
 214                 else
 215                         addr = iph->saddr;
 216
 217                 if ((old_addr ^ addr) & mask)
 218                         break;
 219
 220                 if (skb_try_make_writable(skb, ihl + sizeof(*icmph) +
 221                                           sizeof(*iph) + noff))
 222                         goto drop;
 223
 224                 icmph = (void *)(skb_network_header(skb) + ihl);
 225                 iph = (void *)(icmph + 1);
 226
 227                 new_addr &= mask;
 228                 new_addr |= addr & ~mask;
 229
 230                 /* XXX Fix up the inner checksums. */
 231                 if (egress)
 232                         iph->daddr = new_addr;
 233                 else
 234                         iph->saddr = new_addr;
 235
 236                 inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
 237                                          false);
 238                 break;
 239         }
 240         default:
 241                 break;
 242         }
 243
 244 out:
 245         return action;
 246
 247 drop:
 248         spin_lock(&p->tcf_lock);
 249         p->tcf_qstats.drops++;
 250         spin_unlock(&p->tcf_lock);
 251         return TC_ACT_SHOT;
 252 }
 253
 254 static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
 255                         int bind, int ref)
 256 {
 257         unsigned char *b = skb_tail_pointer(skb);
 258         struct tcf_nat *p = to_tcf_nat(a);
 259         struct tc_nat opt = {
 260                 .old_addr = p->old_addr,
 261                 .new_addr = p->new_addr,
 262                 .mask     = p->mask,
 263                 .flags    = p->flags,
 264
 265                 .index    = p->tcf_index,
 266                 .action   = p->tcf_action,
 267                 .refcnt   = refcount_read(&p->tcf_refcnt) - ref,
 268                 .bindcnt  = atomic_read(&p->tcf_bindcnt) - bind,
 269         };
 270         struct tcf_t t;
 271
 272         if (nla_put(skb, TCA_NAT_PARMS, sizeof(opt), &opt))
 273                 goto nla_put_failure;
 274
 275         tcf_tm_dump(&t, &p->tcf_tm);
 276         if (nla_put_64bit(skb, TCA_NAT_TM, sizeof(t), &t, TCA_NAT_PAD))
 277                 goto nla_put_failure;
 278
 279         return skb->len;
 280
 281 nla_put_failure:
 282         nlmsg_trim(skb, b);
 283         return -1;
 284 }
 285
 286 static int tcf_nat_walker(struct net *net, struct sk_buff *skb,
 287                           struct netlink_callback *cb, int type,
 288                           const struct tc_action_ops *ops,
 289                           struct netlink_ext_ack *extack)
 290 {
 291         struct tc_action_net *tn = net_generic(net, nat_net_id);
 292
 293         return tcf_generic_walker(tn, skb, cb, type, ops, extack);
 294 }
 295
 296 static int tcf_nat_search(struct net *net, struct tc_action **a, u32 index,
 297                           struct netlink_ext_ack *extack)
 298 {
 299         struct tc_action_net *tn = net_generic(net, nat_net_id);
 300
 301         return tcf_idr_search(tn, a, index);
 302 }
 303
 304 static struct tc_action_ops act_nat_ops = {
 305         .kind           =       "nat",
 306         .type           =       TCA_ACT_NAT,
 307         .owner          =       THIS_MODULE,
 308         .act            =       tcf_nat_act,
 309         .dump           =       tcf_nat_dump,
 310         .init           =       tcf_nat_init,
 311         .walk           =       tcf_nat_walker,
 312         .lookup         =       tcf_nat_search,
 313         .size           =       sizeof(struct tcf_nat),
 314 };
 315
 316 static __net_init int nat_init_net(struct net *net)
 317 {
 318         struct tc_action_net *tn = net_generic(net, nat_net_id);
 319
 320         return tc_action_net_init(net, tn, &act_nat_ops);
 321 }
 322
 323 static void __net_exit nat_exit_net(struct list_head *net_list)
 324 {
 325         tc_action_net_exit(net_list, nat_net_id);
 326 }
 327
 328 static struct pernet_operations nat_net_ops = {
 329         .init = nat_init_net,
 330         .exit_batch = nat_exit_net,
 331         .id   = &nat_net_id,
 332         .size = sizeof(struct tc_action_net),
 333 };
 334
 335 MODULE_DESCRIPTION("Stateless NAT actions");
 336 MODULE_LICENSE("GPL");
 337
 338 static int __init nat_init_module(void)
 339 {
 340         return tcf_register_action(&act_nat_ops, &nat_net_ops);
 341 }
 342
 343 static void __exit nat_cleanup_module(void)
 344 {
 345         tcf_unregister_action(&act_nat_ops, &nat_net_ops);
 346 }
 347
 348 module_init(nat_init_module);
 349 module_exit(nat_cleanup_module);