net/sched/act_nat.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * Stateless NAT actions
   4  *
   5  * Copyright (c) 2007 Herbert Xu <herbert@gondor.apana.org.au>
   6  */
   7
   8 #include <linux/errno.h>
   9 #include <linux/init.h>
  10 #include <linux/kernel.h>
  11 #include <linux/module.h>
  12 #include <linux/netfilter.h>
  13 #include <linux/rtnetlink.h>
  14 #include <linux/skbuff.h>
  15 #include <linux/slab.h>
  16 #include <linux/spinlock.h>
  17 #include <linux/string.h>
  18 #include <linux/tc_act/tc_nat.h>
  19 #include <net/act_api.h>
  20 #include <net/pkt_cls.h>
  21 #include <net/icmp.h>
  22 #include <net/ip.h>
  23 #include <net/netlink.h>
  24 #include <net/tc_act/tc_nat.h>
  25 #include <net/tcp.h>
  26 #include <net/udp.h>
  27
  28
  29 static unsigned int nat_net_id;
  30 static struct tc_action_ops act_nat_ops;
  31
  32 static const struct nla_policy nat_policy[TCA_NAT_MAX + 1] = {
  33         [TCA_NAT_PARMS] = { .len = sizeof(struct tc_nat) },
  34 };
  35
  36 static int tcf_nat_init(struct net *net, struct nlattr *nla, struct nlattr *est,
  37                         struct tc_action **a, int ovr, int bind,
  38                         bool rtnl_held, struct tcf_proto *tp,
  39                         u32 flags, struct netlink_ext_ack *extack)
  40 {
  41         struct tc_action_net *tn = net_generic(net, nat_net_id);
  42         struct nlattr *tb[TCA_NAT_MAX + 1];
  43         struct tcf_chain *goto_ch = NULL;
  44         struct tc_nat *parm;
  45         int ret = 0, err;
  46         struct tcf_nat *p;
  47         u32 index;
  48
  49         if (nla == NULL)
  50                 return -EINVAL;
  51
  52         err = nla_parse_nested_deprecated(tb, TCA_NAT_MAX, nla, nat_policy,
  53                                           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, 0);
  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         err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
  81         if (err < 0)
  82                 goto release_idr;
  83         p = to_tcf_nat(*a);
  84
  85         spin_lock_bh(&p->tcf_lock);
  86         p->old_addr = parm->old_addr;
  87         p->new_addr = parm->new_addr;
  88         p->mask = parm->mask;
  89         p->flags = parm->flags;
  90
  91         goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
  92         spin_unlock_bh(&p->tcf_lock);
  93         if (goto_ch)
  94                 tcf_chain_put_by_act(goto_ch);
  95
  96         return ret;
  97 release_idr:
  98         tcf_idr_release(*a, bind);
  99         return err;
 100 }
 101
 102 static int tcf_nat_act(struct sk_buff *skb, const struct tc_action *a,
 103                        struct tcf_result *res)
 104 {
 105         struct tcf_nat *p = to_tcf_nat(a);
 106         struct iphdr *iph;
 107         __be32 old_addr;
 108         __be32 new_addr;
 109         __be32 mask;
 110         __be32 addr;
 111         int egress;
 112         int action;
 113         int ihl;
 114         int noff;
 115
 116         spin_lock(&p->tcf_lock);
 117
 118         tcf_lastuse_update(&p->tcf_tm);
 119         old_addr = p->old_addr;
 120         new_addr = p->new_addr;
 121         mask = p->mask;
 122         egress = p->flags & TCA_NAT_FLAG_EGRESS;
 123         action = p->tcf_action;
 124
 125         bstats_update(&p->tcf_bstats, skb);
 126
 127         spin_unlock(&p->tcf_lock);
 128
 129         if (unlikely(action == TC_ACT_SHOT))
 130                 goto drop;
 131
 132         noff = skb_network_offset(skb);
 133         if (!pskb_may_pull(skb, sizeof(*iph) + noff))
 134                 goto drop;
 135
 136         iph = ip_hdr(skb);
 137
 138         if (egress)
 139                 addr = iph->saddr;
 140         else
 141                 addr = iph->daddr;
 142
 143         if (!((old_addr ^ addr) & mask)) {
 144                 if (skb_try_make_writable(skb, sizeof(*iph) + noff))
 145                         goto drop;
 146
 147                 new_addr &= mask;
 148                 new_addr |= addr & ~mask;
 149
 150                 /* Rewrite IP header */
 151                 iph = ip_hdr(skb);
 152                 if (egress)
 153                         iph->saddr = new_addr;
 154                 else
 155                         iph->daddr = new_addr;
 156
 157                 csum_replace4(&iph->check, addr, new_addr);
 158         } else if ((iph->frag_off & htons(IP_OFFSET)) ||
 159                    iph->protocol != IPPROTO_ICMP) {
 160                 goto out;
 161         }
 162
 163         ihl = iph->ihl * 4;
 164
 165         /* It would be nice to share code with stateful NAT. */
 166         switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
 167         case IPPROTO_TCP:
 168         {
 169                 struct tcphdr *tcph;
 170
 171                 if (!pskb_may_pull(skb, ihl + sizeof(*tcph) + noff) ||
 172                     skb_try_make_writable(skb, ihl + sizeof(*tcph) + noff))
 173                         goto drop;
 174
 175                 tcph = (void *)(skb_network_header(skb) + ihl);
 176                 inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr,
 177                                          true);
 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_try_make_writable(skb, ihl + sizeof(*udph) + noff))
 186                         goto drop;
 187
 188                 udph = (void *)(skb_network_header(skb) + ihl);
 189                 if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
 190                         inet_proto_csum_replace4(&udph->check, skb, addr,
 191                                                  new_addr, true);
 192                         if (!udph->check)
 193                                 udph->check = CSUM_MANGLED_0;
 194                 }
 195                 break;
 196         }
 197         case IPPROTO_ICMP:
 198         {
 199                 struct icmphdr *icmph;
 200
 201                 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + noff))
 202                         goto drop;
 203
 204                 icmph = (void *)(skb_network_header(skb) + ihl);
 205
 206                 if (!icmp_is_err(icmph->type))
 207                         break;
 208
 209                 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph) +
 210                                         noff))
 211                         goto drop;
 212
 213                 icmph = (void *)(skb_network_header(skb) + ihl);
 214                 iph = (void *)(icmph + 1);
 215                 if (egress)
 216                         addr = iph->daddr;
 217                 else
 218                         addr = iph->saddr;
 219
 220                 if ((old_addr ^ addr) & mask)
 221                         break;
 222
 223                 if (skb_try_make_writable(skb, ihl + sizeof(*icmph) +
 224                                           sizeof(*iph) + noff))
 225                         goto drop;
 226
 227                 icmph = (void *)(skb_network_header(skb) + ihl);
 228                 iph = (void *)(icmph + 1);
 229
 230                 new_addr &= mask;
 231                 new_addr |= addr & ~mask;
 232
 233                 /* XXX Fix up the inner checksums. */
 234                 if (egress)
 235                         iph->daddr = new_addr;
 236                 else
 237                         iph->saddr = new_addr;
 238
 239                 inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
 240                                          false);
 241                 break;
 242         }
 243         default:
 244                 break;
 245         }
 246
 247 out:
 248         return action;
 249
 250 drop:
 251         spin_lock(&p->tcf_lock);
 252         p->tcf_qstats.drops++;
 253         spin_unlock(&p->tcf_lock);
 254         return TC_ACT_SHOT;
 255 }
 256
 257 static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
 258                         int bind, int ref)
 259 {
 260         unsigned char *b = skb_tail_pointer(skb);
 261         struct tcf_nat *p = to_tcf_nat(a);
 262         struct tc_nat opt = {
 263                 .index    = p->tcf_index,
 264                 .refcnt   = refcount_read(&p->tcf_refcnt) - ref,
 265                 .bindcnt  = atomic_read(&p->tcf_bindcnt) - bind,
 266         };
 267         struct tcf_t t;
 268
 269         spin_lock_bh(&p->tcf_lock);
 270         opt.old_addr = p->old_addr;
 271         opt.new_addr = p->new_addr;
 272         opt.mask = p->mask;
 273         opt.flags = p->flags;
 274         opt.action = p->tcf_action;
 275
 276         if (nla_put(skb, TCA_NAT_PARMS, sizeof(opt), &opt))
 277                 goto nla_put_failure;
 278
 279         tcf_tm_dump(&t, &p->tcf_tm);
 280         if (nla_put_64bit(skb, TCA_NAT_TM, sizeof(t), &t, TCA_NAT_PAD))
 281                 goto nla_put_failure;
 282         spin_unlock_bh(&p->tcf_lock);
 283
 284         return skb->len;
 285
 286 nla_put_failure:
 287         spin_unlock_bh(&p->tcf_lock);
 288         nlmsg_trim(skb, b);
 289         return -1;
 290 }
 291
 292 static int tcf_nat_walker(struct net *net, struct sk_buff *skb,
 293                           struct netlink_callback *cb, int type,
 294                           const struct tc_action_ops *ops,
 295                           struct netlink_ext_ack *extack)
 296 {
 297         struct tc_action_net *tn = net_generic(net, nat_net_id);
 298
 299         return tcf_generic_walker(tn, skb, cb, type, ops, extack);
 300 }
 301
 302 static int tcf_nat_search(struct net *net, struct tc_action **a, u32 index)
 303 {
 304         struct tc_action_net *tn = net_generic(net, nat_net_id);
 305
 306         return tcf_idr_search(tn, a, index);
 307 }
 308
 309 static struct tc_action_ops act_nat_ops = {
 310         .kind           =       "nat",
 311         .id             =       TCA_ID_NAT,
 312         .owner          =       THIS_MODULE,
 313         .act            =       tcf_nat_act,
 314         .dump           =       tcf_nat_dump,
 315         .init           =       tcf_nat_init,
 316         .walk           =       tcf_nat_walker,
 317         .lookup         =       tcf_nat_search,
 318         .size           =       sizeof(struct tcf_nat),
 319 };
 320
 321 static __net_init int nat_init_net(struct net *net)
 322 {
 323         struct tc_action_net *tn = net_generic(net, nat_net_id);
 324
 325         return tc_action_net_init(net, tn, &act_nat_ops);
 326 }
 327
 328 static void __net_exit nat_exit_net(struct list_head *net_list)
 329 {
 330         tc_action_net_exit(net_list, nat_net_id);
 331 }
 332
 333 static struct pernet_operations nat_net_ops = {
 334         .init = nat_init_net,
 335         .exit_batch = nat_exit_net,
 336         .id   = &nat_net_id,
 337         .size = sizeof(struct tc_action_net),
 338 };
 339
 340 MODULE_DESCRIPTION("Stateless NAT actions");
 341 MODULE_LICENSE("GPL");
 342
 343 static int __init nat_init_module(void)
 344 {
 345         return tcf_register_action(&act_nat_ops, &nat_net_ops);
 346 }
 347
 348 static void __exit nat_cleanup_module(void)
 349 {
 350         tcf_unregister_action(&act_nat_ops, &nat_net_ops);
 351 }
 352
 353 module_init(nat_init_module);
 354 module_exit(nat_cleanup_module);