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