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 <[email protected]>
   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         if (ret == ACT_P_CREATED)
  97                 tcf_idr_insert(tn, *a);
  98
  99         return ret;
 100 release_idr:
 101         tcf_idr_release(*a, bind);
 102         return err;
 103 }
 104
 105 static int tcf_nat_act(struct sk_buff *skb, const struct tc_action *a,
 106                        struct tcf_result *res)
 107 {
 108         struct tcf_nat *p = to_tcf_nat(a);
 109         struct iphdr *iph;
 110         __be32 old_addr;
 111         __be32 new_addr;
 112         __be32 mask;
 113         __be32 addr;
 114         int egress;
 115         int action;
 116         int ihl;
 117         int noff;
 118
 119         spin_lock(&p->tcf_lock);
 120
 121         tcf_lastuse_update(&p->tcf_tm);
 122         old_addr = p->old_addr;
 123         new_addr = p->new_addr;
 124         mask = p->mask;
 125         egress = p->flags & TCA_NAT_FLAG_EGRESS;
 126         action = p->tcf_action;
 127
 128         bstats_update(&p->tcf_bstats, skb);
 129
 130         spin_unlock(&p->tcf_lock);
 131
 132         if (unlikely(action == TC_ACT_SHOT))
 133                 goto drop;
 134
 135         noff = skb_network_offset(skb);
 136         if (!pskb_may_pull(skb, sizeof(*iph) + noff))
 137                 goto drop;
 138
 139         iph = ip_hdr(skb);
 140
 141         if (egress)
 142                 addr = iph->saddr;
 143         else
 144                 addr = iph->daddr;
 145
 146         if (!((old_addr ^ addr) & mask)) {
 147                 if (skb_try_make_writable(skb, sizeof(*iph) + noff))
 148                         goto drop;
 149
 150                 new_addr &= mask;
 151                 new_addr |= addr & ~mask;
 152
 153                 /* Rewrite IP header */
 154                 iph = ip_hdr(skb);
 155                 if (egress)
 156                         iph->saddr = new_addr;
 157                 else
 158                         iph->daddr = new_addr;
 159
 160                 csum_replace4(&iph->check, addr, new_addr);
 161         } else if ((iph->frag_off & htons(IP_OFFSET)) ||
 162                    iph->protocol != IPPROTO_ICMP) {
 163                 goto out;
 164         }
 165
 166         ihl = iph->ihl * 4;
 167
 168         /* It would be nice to share code with stateful NAT. */
 169         switch (iph->frag_off & htons(IP_OFFSET) ? 0 : iph->protocol) {
 170         case IPPROTO_TCP:
 171         {
 172                 struct tcphdr *tcph;
 173
 174                 if (!pskb_may_pull(skb, ihl + sizeof(*tcph) + noff) ||
 175                     skb_try_make_writable(skb, ihl + sizeof(*tcph) + noff))
 176                         goto drop;
 177
 178                 tcph = (void *)(skb_network_header(skb) + ihl);
 179                 inet_proto_csum_replace4(&tcph->check, skb, addr, new_addr,
 180                                          true);
 181                 break;
 182         }
 183         case IPPROTO_UDP:
 184         {
 185                 struct udphdr *udph;
 186
 187                 if (!pskb_may_pull(skb, ihl + sizeof(*udph) + noff) ||
 188                     skb_try_make_writable(skb, ihl + sizeof(*udph) + noff))
 189                         goto drop;
 190
 191                 udph = (void *)(skb_network_header(skb) + ihl);
 192                 if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
 193                         inet_proto_csum_replace4(&udph->check, skb, addr,
 194                                                  new_addr, true);
 195                         if (!udph->check)
 196                                 udph->check = CSUM_MANGLED_0;
 197                 }
 198                 break;
 199         }
 200         case IPPROTO_ICMP:
 201         {
 202                 struct icmphdr *icmph;
 203
 204                 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + noff))
 205                         goto drop;
 206
 207                 icmph = (void *)(skb_network_header(skb) + ihl);
 208
 209                 if ((icmph->type != ICMP_DEST_UNREACH) &&
 210                     (icmph->type != ICMP_TIME_EXCEEDED) &&
 211                     (icmph->type != ICMP_PARAMETERPROB))
 212                         break;
 213
 214                 if (!pskb_may_pull(skb, ihl + sizeof(*icmph) + sizeof(*iph) +
 215                                         noff))
 216                         goto drop;
 217
 218                 icmph = (void *)(skb_network_header(skb) + ihl);
 219                 iph = (void *)(icmph + 1);
 220                 if (egress)
 221                         addr = iph->daddr;
 222                 else
 223                         addr = iph->saddr;
 224
 225                 if ((old_addr ^ addr) & mask)
 226                         break;
 227
 228                 if (skb_try_make_writable(skb, ihl + sizeof(*icmph) +
 229                                           sizeof(*iph) + noff))
 230                         goto drop;
 231
 232                 icmph = (void *)(skb_network_header(skb) + ihl);
 233                 iph = (void *)(icmph + 1);
 234
 235                 new_addr &= mask;
 236                 new_addr |= addr & ~mask;
 237
 238                 /* XXX Fix up the inner checksums. */
 239                 if (egress)
 240                         iph->daddr = new_addr;
 241                 else
 242                         iph->saddr = new_addr;
 243
 244                 inet_proto_csum_replace4(&icmph->checksum, skb, addr, new_addr,
 245                                          false);
 246                 break;
 247         }
 248         default:
 249                 break;
 250         }
 251
 252 out:
 253         return action;
 254
 255 drop:
 256         spin_lock(&p->tcf_lock);
 257         p->tcf_qstats.drops++;
 258         spin_unlock(&p->tcf_lock);
 259         return TC_ACT_SHOT;
 260 }
 261
 262 static int tcf_nat_dump(struct sk_buff *skb, struct tc_action *a,
 263                         int bind, int ref)
 264 {
 265         unsigned char *b = skb_tail_pointer(skb);
 266         struct tcf_nat *p = to_tcf_nat(a);
 267         struct tc_nat opt = {
 268                 .index    = p->tcf_index,
 269                 .refcnt   = refcount_read(&p->tcf_refcnt) - ref,
 270                 .bindcnt  = atomic_read(&p->tcf_bindcnt) - bind,
 271         };
 272         struct tcf_t t;
 273
 274         spin_lock_bh(&p->tcf_lock);
 275         opt.old_addr = p->old_addr;
 276         opt.new_addr = p->new_addr;
 277         opt.mask = p->mask;
 278         opt.flags = p->flags;
 279         opt.action = p->tcf_action;
 280
 281         if (nla_put(skb, TCA_NAT_PARMS, sizeof(opt), &opt))
 282                 goto nla_put_failure;
 283
 284         tcf_tm_dump(&t, &p->tcf_tm);
 285         if (nla_put_64bit(skb, TCA_NAT_TM, sizeof(t), &t, TCA_NAT_PAD))
 286                 goto nla_put_failure;
 287         spin_unlock_bh(&p->tcf_lock);
 288
 289         return skb->len;
 290
 291 nla_put_failure:
 292         spin_unlock_bh(&p->tcf_lock);
 293         nlmsg_trim(skb, b);
 294         return -1;
 295 }
 296
 297 static int tcf_nat_walker(struct net *net, struct sk_buff *skb,
 298                           struct netlink_callback *cb, int type,
 299                           const struct tc_action_ops *ops,
 300                           struct netlink_ext_ack *extack)
 301 {
 302         struct tc_action_net *tn = net_generic(net, nat_net_id);
 303
 304         return tcf_generic_walker(tn, skb, cb, type, ops, extack);
 305 }
 306
 307 static int tcf_nat_search(struct net *net, struct tc_action **a, u32 index)
 308 {
 309         struct tc_action_net *tn = net_generic(net, nat_net_id);
 310
 311         return tcf_idr_search(tn, a, index);
 312 }
 313
 314 static struct tc_action_ops act_nat_ops = {
 315         .kind           =       "nat",
 316         .id             =       TCA_ID_NAT,
 317         .owner          =       THIS_MODULE,
 318         .act            =       tcf_nat_act,
 319         .dump           =       tcf_nat_dump,
 320         .init           =       tcf_nat_init,
 321         .walk           =       tcf_nat_walker,
 322         .lookup         =       tcf_nat_search,
 323         .size           =       sizeof(struct tcf_nat),
 324 };
 325
 326 static __net_init int nat_init_net(struct net *net)
 327 {
 328         struct tc_action_net *tn = net_generic(net, nat_net_id);
 329
 330         return tc_action_net_init(net, tn, &act_nat_ops);
 331 }
 332
 333 static void __net_exit nat_exit_net(struct list_head *net_list)
 334 {
 335         tc_action_net_exit(net_list, nat_net_id);
 336 }
 337
 338 static struct pernet_operations nat_net_ops = {
 339         .init = nat_init_net,
 340         .exit_batch = nat_exit_net,
 341         .id   = &nat_net_id,
 342         .size = sizeof(struct tc_action_net),
 343 };
 344
 345 MODULE_DESCRIPTION("Stateless NAT actions");
 346 MODULE_LICENSE("GPL");
 347
 348 static int __init nat_init_module(void)
 349 {
 350         return tcf_register_action(&act_nat_ops, &nat_net_ops);
 351 }
 352
 353 static void __exit nat_cleanup_module(void)
 354 {
 355         tcf_unregister_action(&act_nat_ops, &nat_net_ops);
 356 }
 357
 358 module_init(nat_init_module);
 359 module_exit(nat_cleanup_module);