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