net/sched/act_ipt.c

   1 /*
   2  * net/sched/ipt.c      iptables target interface
   3  *
   4  *TODO: Add other tables. For now we only support the ipv4 table targets
   5  *
   6  *              This program is free software; you can redistribute it and/or
   7  *              modify it under the terms of the GNU General Public License
   8  *              as published by the Free Software Foundation; either version
   9  *              2 of the License, or (at your option) any later version.
  10  *
  11  * Copyright:   Jamal Hadi Salim (2002-4)
  12  */
  13
  14 #include <asm/uaccess.h>
  15 #include <asm/system.h>
  16 #include <asm/bitops.h>
  17 #include <linux/types.h>
  18 #include <linux/kernel.h>
  19 #include <linux/sched.h>
  20 #include <linux/string.h>
  21 #include <linux/mm.h>
  22 #include <linux/socket.h>
  23 #include <linux/sockios.h>
  24 #include <linux/in.h>
  25 #include <linux/errno.h>
  26 #include <linux/interrupt.h>
  27 #include <linux/netdevice.h>
  28 #include <linux/skbuff.h>
  29 #include <linux/rtnetlink.h>
  30 #include <linux/module.h>
  31 #include <linux/init.h>
  32 #include <linux/proc_fs.h>
  33 #include <linux/kmod.h>
  34 #include <net/sock.h>
  35 #include <net/pkt_sched.h>
  36 #include <linux/tc_act/tc_ipt.h>
  37 #include <net/tc_act/tc_ipt.h>
  38
  39 #include <linux/netfilter_ipv4/ip_tables.h>
  40
  41
  42 #define IPT_TAB_MASK     15
  43 static struct tcf_common *tcf_ipt_ht[IPT_TAB_MASK + 1];
  44 static u32 ipt_idx_gen;
  45 static DEFINE_RWLOCK(ipt_lock);
  46
  47 static struct tcf_hashinfo ipt_hash_info = {
  48         .htab   =       tcf_ipt_ht,
  49         .hmask  =       IPT_TAB_MASK,
  50         .lock   =       &ipt_lock,
  51 };
  52
  53 static int ipt_init_target(struct ipt_entry_target *t, char *table, unsigned int hook)
  54 {
  55         struct ipt_target *target;
  56         int ret = 0;
  57
  58         target = xt_find_target(AF_INET, t->u.user.name, t->u.user.revision);
  59         if (!target)
  60                 return -ENOENT;
  61
  62         t->u.kernel.target = target;
  63
  64         ret = xt_check_target(target, AF_INET, t->u.target_size - sizeof(*t),
  65                               table, hook, 0, 0);
  66         if (ret)
  67                 return ret;
  68
  69         if (t->u.kernel.target->checkentry
  70             && !t->u.kernel.target->checkentry(table, NULL,
  71                                                t->u.kernel.target, t->data,
  72                                                hook)) {
  73                 module_put(t->u.kernel.target->me);
  74                 ret = -EINVAL;
  75         }
  76
  77         return ret;
  78 }
  79
  80 static void ipt_destroy_target(struct ipt_entry_target *t)
  81 {
  82         if (t->u.kernel.target->destroy)
  83                 t->u.kernel.target->destroy(t->u.kernel.target, t->data);
  84         module_put(t->u.kernel.target->me);
  85 }
  86
  87 static int tcf_ipt_release(struct tcf_ipt *ipt, int bind)
  88 {
  89         int ret = 0;
  90         if (ipt) {
  91                 if (bind)
  92                         ipt->tcf_bindcnt--;
  93                 ipt->tcf_refcnt--;
  94                 if (ipt->tcf_bindcnt <= 0 && ipt->tcf_refcnt <= 0) {
  95                         ipt_destroy_target(ipt->tcfi_t);
  96                         kfree(ipt->tcfi_tname);
  97                         kfree(ipt->tcfi_t);
  98                         tcf_hash_destroy(&ipt->common, &ipt_hash_info);
  99                         ret = ACT_P_DELETED;
 100                 }
 101         }
 102         return ret;
 103 }
 104
 105 static int tcf_ipt_init(struct rtattr *rta, struct rtattr *est,
 106                         struct tc_action *a, int ovr, int bind)
 107 {
 108         struct rtattr *tb[TCA_IPT_MAX];
 109         struct tcf_ipt *ipt;
 110         struct tcf_common *pc;
 111         struct ipt_entry_target *td, *t;
 112         char *tname;
 113         int ret = 0, err;
 114         u32 hook = 0;
 115         u32 index = 0;
 116
 117         if (rta == NULL || rtattr_parse_nested(tb, TCA_IPT_MAX, rta) < 0)
 118                 return -EINVAL;
 119
 120         if (tb[TCA_IPT_HOOK-1] == NULL ||
 121             RTA_PAYLOAD(tb[TCA_IPT_HOOK-1]) < sizeof(u32))
 122                 return -EINVAL;
 123         if (tb[TCA_IPT_TARG-1] == NULL ||
 124             RTA_PAYLOAD(tb[TCA_IPT_TARG-1]) < sizeof(*t))
 125                 return -EINVAL;
 126         td = (struct ipt_entry_target *)RTA_DATA(tb[TCA_IPT_TARG-1]);
 127         if (RTA_PAYLOAD(tb[TCA_IPT_TARG-1]) < td->u.target_size)
 128                 return -EINVAL;
 129
 130         if (tb[TCA_IPT_INDEX-1] != NULL &&
 131             RTA_PAYLOAD(tb[TCA_IPT_INDEX-1]) >= sizeof(u32))
 132                 index = *(u32 *)RTA_DATA(tb[TCA_IPT_INDEX-1]);
 133
 134         pc = tcf_hash_check(index, a, bind, &ipt_hash_info);
 135         if (!pc) {
 136                 pc = tcf_hash_create(index, est, a, sizeof(*ipt), bind,
 137                                      &ipt_idx_gen, &ipt_hash_info);
 138                 if (unlikely(!pc))
 139                         return -ENOMEM;
 140                 ret = ACT_P_CREATED;
 141         } else {
 142                 if (!ovr) {
 143                         tcf_ipt_release(to_ipt(pc), bind);
 144                         return -EEXIST;
 145                 }
 146         }
 147         ipt = to_ipt(pc);
 148
 149         hook = *(u32 *)RTA_DATA(tb[TCA_IPT_HOOK-1]);
 150
 151         err = -ENOMEM;
 152         tname = kmalloc(IFNAMSIZ, GFP_KERNEL);
 153         if (unlikely(!tname))
 154                 goto err1;
 155         if (tb[TCA_IPT_TABLE - 1] == NULL ||
 156             rtattr_strlcpy(tname, tb[TCA_IPT_TABLE-1], IFNAMSIZ) >= IFNAMSIZ)
 157                 strcpy(tname, "mangle");
 158
 159         t = kmemdup(td, td->u.target_size, GFP_KERNEL);
 160         if (unlikely(!t))
 161                 goto err2;
 162
 163         if ((err = ipt_init_target(t, tname, hook)) < 0)
 164                 goto err3;
 165
 166         spin_lock_bh(&ipt->tcf_lock);
 167         if (ret != ACT_P_CREATED) {
 168                 ipt_destroy_target(ipt->tcfi_t);
 169                 kfree(ipt->tcfi_tname);
 170                 kfree(ipt->tcfi_t);
 171         }
 172         ipt->tcfi_tname = tname;
 173         ipt->tcfi_t     = t;
 174         ipt->tcfi_hook  = hook;
 175         spin_unlock_bh(&ipt->tcf_lock);
 176         if (ret == ACT_P_CREATED)
 177                 tcf_hash_insert(pc, &ipt_hash_info);
 178         return ret;
 179
 180 err3:
 181         kfree(t);
 182 err2:
 183         kfree(tname);
 184 err1:
 185         kfree(pc);
 186         return err;
 187 }
 188
 189 static int tcf_ipt_cleanup(struct tc_action *a, int bind)
 190 {
 191         struct tcf_ipt *ipt = a->priv;
 192         return tcf_ipt_release(ipt, bind);
 193 }
 194
 195 static int tcf_ipt(struct sk_buff *skb, struct tc_action *a,
 196                    struct tcf_result *res)
 197 {
 198         int ret = 0, result = 0;
 199         struct tcf_ipt *ipt = a->priv;
 200
 201         if (skb_cloned(skb)) {
 202                 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC))
 203                         return TC_ACT_UNSPEC;
 204         }
 205
 206         spin_lock(&ipt->tcf_lock);
 207
 208         ipt->tcf_tm.lastuse = jiffies;
 209         ipt->tcf_bstats.bytes += skb->len;
 210         ipt->tcf_bstats.packets++;
 211
 212         /* yes, we have to worry about both in and out dev
 213          worry later - danger - this API seems to have changed
 214          from earlier kernels */
 215
 216         /* iptables targets take a double skb pointer in case the skb
 217          * needs to be replaced. We don't own the skb, so this must not
 218          * happen. The pskb_expand_head above should make sure of this */
 219         ret = ipt->tcfi_t->u.kernel.target->target(&skb, skb->dev, NULL,
 220                                                    ipt->tcfi_hook,
 221                                                    ipt->tcfi_t->u.kernel.target,
 222                                                    ipt->tcfi_t->data);
 223         switch (ret) {
 224         case NF_ACCEPT:
 225                 result = TC_ACT_OK;
 226                 break;
 227         case NF_DROP:
 228                 result = TC_ACT_SHOT;
 229                 ipt->tcf_qstats.drops++;
 230                 break;
 231         case IPT_CONTINUE:
 232                 result = TC_ACT_PIPE;
 233                 break;
 234         default:
 235                 if (net_ratelimit())
 236                         printk("Bogus netfilter code %d assume ACCEPT\n", ret);
 237                 result = TC_POLICE_OK;
 238                 break;
 239         }
 240         spin_unlock(&ipt->tcf_lock);
 241         return result;
 242
 243 }
 244
 245 static int tcf_ipt_dump(struct sk_buff *skb, struct tc_action *a, int bind, int ref)
 246 {
 247         unsigned char *b = skb->tail;
 248         struct tcf_ipt *ipt = a->priv;
 249         struct ipt_entry_target *t;
 250         struct tcf_t tm;
 251         struct tc_cnt c;
 252
 253         /* for simple targets kernel size == user size
 254         ** user name = target name
 255         ** for foolproof you need to not assume this
 256         */
 257
 258         t = kmemdup(ipt->tcfi_t, ipt->tcfi_t->u.user.target_size, GFP_ATOMIC);
 259         if (unlikely(!t))
 260                 goto rtattr_failure;
 261
 262         c.bindcnt = ipt->tcf_bindcnt - bind;
 263         c.refcnt = ipt->tcf_refcnt - ref;
 264         strcpy(t->u.user.name, ipt->tcfi_t->u.kernel.target->name);
 265
 266         RTA_PUT(skb, TCA_IPT_TARG, ipt->tcfi_t->u.user.target_size, t);
 267         RTA_PUT(skb, TCA_IPT_INDEX, 4, &ipt->tcf_index);
 268         RTA_PUT(skb, TCA_IPT_HOOK, 4, &ipt->tcfi_hook);
 269         RTA_PUT(skb, TCA_IPT_CNT, sizeof(struct tc_cnt), &c);
 270         RTA_PUT(skb, TCA_IPT_TABLE, IFNAMSIZ, ipt->tcfi_tname);
 271         tm.install = jiffies_to_clock_t(jiffies - ipt->tcf_tm.install);
 272         tm.lastuse = jiffies_to_clock_t(jiffies - ipt->tcf_tm.lastuse);
 273         tm.expires = jiffies_to_clock_t(ipt->tcf_tm.expires);
 274         RTA_PUT(skb, TCA_IPT_TM, sizeof (tm), &tm);
 275         kfree(t);
 276         return skb->len;
 277
 278 rtattr_failure:
 279         skb_trim(skb, b - skb->data);
 280         kfree(t);
 281         return -1;
 282 }
 283
 284 static struct tc_action_ops act_ipt_ops = {
 285         .kind           =       "ipt",
 286         .hinfo          =       &ipt_hash_info,
 287         .type           =       TCA_ACT_IPT,
 288         .capab          =       TCA_CAP_NONE,
 289         .owner          =       THIS_MODULE,
 290         .act            =       tcf_ipt,
 291         .dump           =       tcf_ipt_dump,
 292         .cleanup        =       tcf_ipt_cleanup,
 293         .lookup         =       tcf_hash_search,
 294         .init           =       tcf_ipt_init,
 295         .walk           =       tcf_generic_walker
 296 };
 297
 298 MODULE_AUTHOR("Jamal Hadi Salim(2002-4)");
 299 MODULE_DESCRIPTION("Iptables target actions");
 300 MODULE_LICENSE("GPL");
 301
 302 static int __init ipt_init_module(void)
 303 {
 304         return tcf_register_action(&act_ipt_ops);
 305 }
 306
 307 static void __exit ipt_cleanup_module(void)
 308 {
 309         tcf_unregister_action(&act_ipt_ops);
 310 }
 311
 312 module_init(ipt_init_module);
 313 module_exit(ipt_cleanup_module);