net/netfilter/core.c

   1 /* netfilter.c: look after the filters for various protocols.
   2  * Heavily influenced by the old firewall.c by David Bonn and Alan Cox.
   3  *
   4  * Thanks to Rob `CmdrTaco' Malda for not influencing this code in any
   5  * way.
   6  *
   7  * Rusty Russell (C)2000 -- This code is GPL.
   8  * Patrick McHardy (c) 2006-2012
   9  */
  10 #include <linux/kernel.h>
  11 #include <linux/netfilter.h>
  12 #include <net/protocol.h>
  13 #include <linux/init.h>
  14 #include <linux/skbuff.h>
  15 #include <linux/wait.h>
  16 #include <linux/module.h>
  17 #include <linux/interrupt.h>
  18 #include <linux/if.h>
  19 #include <linux/netdevice.h>
  20 #include <linux/netfilter_ipv6.h>
  21 #include <linux/inetdevice.h>
  22 #include <linux/proc_fs.h>
  23 #include <linux/mutex.h>
  24 #include <linux/slab.h>
  25 #include <net/net_namespace.h>
  26 #include <net/sock.h>
  27
  28 #include "nf_internals.h"
  29
  30 static DEFINE_MUTEX(afinfo_mutex);
  31
  32 const struct nf_afinfo __rcu *nf_afinfo[NFPROTO_NUMPROTO] __read_mostly;
  33 EXPORT_SYMBOL(nf_afinfo);
  34 const struct nf_ipv6_ops __rcu *nf_ipv6_ops __read_mostly;
  35 EXPORT_SYMBOL_GPL(nf_ipv6_ops);
  36
  37 int nf_register_afinfo(const struct nf_afinfo *afinfo)
  38 {
  39         mutex_lock(&afinfo_mutex);
  40         RCU_INIT_POINTER(nf_afinfo[afinfo->family], afinfo);
  41         mutex_unlock(&afinfo_mutex);
  42         return 0;
  43 }
  44 EXPORT_SYMBOL_GPL(nf_register_afinfo);
  45
  46 void nf_unregister_afinfo(const struct nf_afinfo *afinfo)
  47 {
  48         mutex_lock(&afinfo_mutex);
  49         RCU_INIT_POINTER(nf_afinfo[afinfo->family], NULL);
  50         mutex_unlock(&afinfo_mutex);
  51         synchronize_rcu();
  52 }
  53 EXPORT_SYMBOL_GPL(nf_unregister_afinfo);
  54
  55 struct list_head nf_hooks[NFPROTO_NUMPROTO][NF_MAX_HOOKS] __read_mostly;
  56 EXPORT_SYMBOL(nf_hooks);
  57
  58 #ifdef HAVE_JUMP_LABEL
  59 struct static_key nf_hooks_needed[NFPROTO_NUMPROTO][NF_MAX_HOOKS];
  60 EXPORT_SYMBOL(nf_hooks_needed);
  61 #endif
  62
  63 static DEFINE_MUTEX(nf_hook_mutex);
  64
  65 int nf_register_hook(struct nf_hook_ops *reg)
  66 {
  67         struct nf_hook_ops *elem;
  68
  69         mutex_lock(&nf_hook_mutex);
  70         list_for_each_entry(elem, &nf_hooks[reg->pf][reg->hooknum], list) {
  71                 if (reg->priority < elem->priority)
  72                         break;
  73         }
  74         list_add_rcu(&reg->list, elem->list.prev);
  75         mutex_unlock(&nf_hook_mutex);
  76 #ifdef HAVE_JUMP_LABEL
  77         static_key_slow_inc(&nf_hooks_needed[reg->pf][reg->hooknum]);
  78 #endif
  79         return 0;
  80 }
  81 EXPORT_SYMBOL(nf_register_hook);
  82
  83 void nf_unregister_hook(struct nf_hook_ops *reg)
  84 {
  85         mutex_lock(&nf_hook_mutex);
  86         list_del_rcu(&reg->list);
  87         mutex_unlock(&nf_hook_mutex);
  88 #ifdef HAVE_JUMP_LABEL
  89         static_key_slow_dec(&nf_hooks_needed[reg->pf][reg->hooknum]);
  90 #endif
  91         synchronize_net();
  92 }
  93 EXPORT_SYMBOL(nf_unregister_hook);
  94
  95 int nf_register_hooks(struct nf_hook_ops *reg, unsigned int n)
  96 {
  97         unsigned int i;
  98         int err = 0;
  99
 100         for (i = 0; i < n; i++) {
 101                 err = nf_register_hook(&reg[i]);
 102                 if (err)
 103                         goto err;
 104         }
 105         return err;
 106
 107 err:
 108         if (i > 0)
 109                 nf_unregister_hooks(reg, i);
 110         return err;
 111 }
 112 EXPORT_SYMBOL(nf_register_hooks);
 113
 114 void nf_unregister_hooks(struct nf_hook_ops *reg, unsigned int n)
 115 {
 116         while (n-- > 0)
 117                 nf_unregister_hook(&reg[n]);
 118 }
 119 EXPORT_SYMBOL(nf_unregister_hooks);
 120
 121 unsigned int nf_iterate(struct list_head *head,
 122                         struct sk_buff *skb,
 123                         struct nf_hook_state *state,
 124                         struct nf_hook_ops **elemp)
 125 {
 126         unsigned int verdict;
 127
 128         /*
 129          * The caller must not block between calls to this
 130          * function because of risk of continuing from deleted element.
 131          */
 132         list_for_each_entry_continue_rcu((*elemp), head, list) {
 133                 if (state->thresh > (*elemp)->priority)
 134                         continue;
 135
 136                 /* Optimization: we don't need to hold module
 137                    reference here, since function can't sleep. --RR */
 138 repeat:
 139                 verdict = (*elemp)->hook(*elemp, skb, state);
 140                 if (verdict != NF_ACCEPT) {
 141 #ifdef CONFIG_NETFILTER_DEBUG
 142                         if (unlikely((verdict & NF_VERDICT_MASK)
 143                                                         > NF_MAX_VERDICT)) {
 144                                 NFDEBUG("Evil return from %p(%u).\n",
 145                                         (*elemp)->hook, state->hook);
 146                                 continue;
 147                         }
 148 #endif
 149                         if (verdict != NF_REPEAT)
 150                                 return verdict;
 151                         goto repeat;
 152                 }
 153         }
 154         return NF_ACCEPT;
 155 }
 156
 157
 158 /* Returns 1 if okfn() needs to be executed by the caller,
 159  * -EPERM for NF_DROP, 0 otherwise. */
 160 int nf_hook_slow(struct sk_buff *skb, struct nf_hook_state *state)
 161 {
 162         struct nf_hook_ops *elem;
 163         unsigned int verdict;
 164         int ret = 0;
 165
 166         /* We may already have this, but read-locks nest anyway */
 167         rcu_read_lock();
 168
 169         elem = list_entry_rcu(state->hook_list, struct nf_hook_ops, list);
 170 next_hook:
 171         verdict = nf_iterate(state->hook_list, skb, state, &elem);
 172         if (verdict == NF_ACCEPT || verdict == NF_STOP) {
 173                 ret = 1;
 174         } else if ((verdict & NF_VERDICT_MASK) == NF_DROP) {
 175                 kfree_skb(skb);
 176                 ret = NF_DROP_GETERR(verdict);
 177                 if (ret == 0)
 178                         ret = -EPERM;
 179         } else if ((verdict & NF_VERDICT_MASK) == NF_QUEUE) {
 180                 int err = nf_queue(skb, elem, state,
 181                                    verdict >> NF_VERDICT_QBITS);
 182                 if (err < 0) {
 183                         if (err == -ECANCELED)
 184                                 goto next_hook;
 185                         if (err == -ESRCH &&
 186                            (verdict & NF_VERDICT_FLAG_QUEUE_BYPASS))
 187                                 goto next_hook;
 188                         kfree_skb(skb);
 189                 }
 190         }
 191         rcu_read_unlock();
 192         return ret;
 193 }
 194 EXPORT_SYMBOL(nf_hook_slow);
 195
 196
 197 int skb_make_writable(struct sk_buff *skb, unsigned int writable_len)
 198 {
 199         if (writable_len > skb->len)
 200                 return 0;
 201
 202         /* Not exclusive use of packet?  Must copy. */
 203         if (!skb_cloned(skb)) {
 204                 if (writable_len <= skb_headlen(skb))
 205                         return 1;
 206         } else if (skb_clone_writable(skb, writable_len))
 207                 return 1;
 208
 209         if (writable_len <= skb_headlen(skb))
 210                 writable_len = 0;
 211         else
 212                 writable_len -= skb_headlen(skb);
 213
 214         return !!__pskb_pull_tail(skb, writable_len);
 215 }
 216 EXPORT_SYMBOL(skb_make_writable);
 217
 218 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
 219 /* This does not belong here, but locally generated errors need it if connection
 220    tracking in use: without this, connection may not be in hash table, and hence
 221    manufactured ICMP or RST packets will not be associated with it. */
 222 void (*ip_ct_attach)(struct sk_buff *, const struct sk_buff *)
 223                 __rcu __read_mostly;
 224 EXPORT_SYMBOL(ip_ct_attach);
 225
 226 void nf_ct_attach(struct sk_buff *new, const struct sk_buff *skb)
 227 {
 228         void (*attach)(struct sk_buff *, const struct sk_buff *);
 229
 230         if (skb->nfct) {
 231                 rcu_read_lock();
 232                 attach = rcu_dereference(ip_ct_attach);
 233                 if (attach)
 234                         attach(new, skb);
 235                 rcu_read_unlock();
 236         }
 237 }
 238 EXPORT_SYMBOL(nf_ct_attach);
 239
 240 void (*nf_ct_destroy)(struct nf_conntrack *) __rcu __read_mostly;
 241 EXPORT_SYMBOL(nf_ct_destroy);
 242
 243 void nf_conntrack_destroy(struct nf_conntrack *nfct)
 244 {
 245         void (*destroy)(struct nf_conntrack *);
 246
 247         rcu_read_lock();
 248         destroy = rcu_dereference(nf_ct_destroy);
 249         BUG_ON(destroy == NULL);
 250         destroy(nfct);
 251         rcu_read_unlock();
 252 }
 253 EXPORT_SYMBOL(nf_conntrack_destroy);
 254
 255 struct nfq_ct_hook __rcu *nfq_ct_hook __read_mostly;
 256 EXPORT_SYMBOL_GPL(nfq_ct_hook);
 257
 258 struct nfq_ct_nat_hook __rcu *nfq_ct_nat_hook __read_mostly;
 259 EXPORT_SYMBOL_GPL(nfq_ct_nat_hook);
 260
 261 #endif /* CONFIG_NF_CONNTRACK */
 262
 263 #ifdef CONFIG_NF_NAT_NEEDED
 264 void (*nf_nat_decode_session_hook)(struct sk_buff *, struct flowi *);
 265 EXPORT_SYMBOL(nf_nat_decode_session_hook);
 266 #endif
 267
 268 static int __net_init netfilter_net_init(struct net *net)
 269 {
 270 #ifdef CONFIG_PROC_FS
 271         net->nf.proc_netfilter = proc_net_mkdir(net, "netfilter",
 272                                                 net->proc_net);
 273         if (!net->nf.proc_netfilter) {
 274                 if (!net_eq(net, &init_net))
 275                         pr_err("cannot create netfilter proc entry");
 276
 277                 return -ENOMEM;
 278         }
 279 #endif
 280         return 0;
 281 }
 282
 283 static void __net_exit netfilter_net_exit(struct net *net)
 284 {
 285         remove_proc_entry("netfilter", net->proc_net);
 286 }
 287
 288 static struct pernet_operations netfilter_net_ops = {
 289         .init = netfilter_net_init,
 290         .exit = netfilter_net_exit,
 291 };
 292
 293 int __init netfilter_init(void)
 294 {
 295         int i, h, ret;
 296
 297         for (i = 0; i < ARRAY_SIZE(nf_hooks); i++) {
 298                 for (h = 0; h < NF_MAX_HOOKS; h++)
 299                         INIT_LIST_HEAD(&nf_hooks[i][h]);
 300         }
 301
 302         ret = register_pernet_subsys(&netfilter_net_ops);
 303         if (ret < 0)
 304                 goto err;
 305
 306         ret = netfilter_log_init();
 307         if (ret < 0)
 308                 goto err_pernet;
 309
 310         return 0;
 311 err_pernet:
 312         unregister_pernet_subsys(&netfilter_net_ops);
 313 err:
 314         return ret;
 315 }