net/sched/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/config.h>
  18 #include <linux/types.h>
  19 #include <linux/kernel.h>
  20 #include <linux/sched.h>
  21 #include <linux/string.h>
  22 #include <linux/mm.h>
  23 #include <linux/socket.h>
  24 #include <linux/sockios.h>
  25 #include <linux/in.h>
  26 #include <linux/errno.h>
  27 #include <linux/interrupt.h>
  28 #include <linux/netdevice.h>
  29 #include <linux/skbuff.h>
  30 #include <linux/rtnetlink.h>
  31 #include <linux/module.h>
  32 #include <linux/init.h>
  33 #include <linux/proc_fs.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 /* use generic hash table */
  42 #define MY_TAB_SIZE     16
  43 #define MY_TAB_MASK     15
  44
  45 static u32 idx_gen;
  46 static struct tcf_ipt *tcf_ipt_ht[MY_TAB_SIZE];
  47 /* ipt hash table lock */
  48 static rwlock_t ipt_lock = RW_LOCK_UNLOCKED;
  49
  50 /* ovewrride the defaults */
  51 #define tcf_st  tcf_ipt
  52 #define tcf_t_lock   ipt_lock
  53 #define tcf_ht tcf_ipt_ht
  54
  55 #include <net/pkt_act.h>
  56
  57 static inline int
  58 init_targ(struct tcf_ipt *p)
  59 {
  60         struct ipt_target *target;
  61         int ret = 0;
  62         struct ipt_entry_target *t = p->t;
  63         target = __ipt_find_target_lock(t->u.user.name, &ret);
  64
  65         if (!target) {
  66                 printk("init_targ: Failed to find %s\n", t->u.user.name);
  67                 return -1;
  68         }
  69
  70         DPRINTK("init_targ: found %s\n", target->name);
  71         /* we really need proper ref counting
  72          seems to be only needed for modules?? Talk to laforge */
  73 /*      if (target->me)
  74               __MOD_INC_USE_COUNT(target->me);
  75 */
  76         t->u.kernel.target = target;
  77
  78         __ipt_mutex_up();
  79
  80         if (t->u.kernel.target->checkentry
  81             && !t->u.kernel.target->checkentry(p->tname, NULL, t->data,
  82                                                t->u.target_size
  83                                                - sizeof (*t), p->hook)) {
  84 /*              if (t->u.kernel.target->me)
  85               __MOD_DEC_USE_COUNT(t->u.kernel.target->me);
  86 */
  87                 DPRINTK("ip_tables: check failed for `%s'.\n",
  88                         t->u.kernel.target->name);
  89                 ret = -EINVAL;
  90         }
  91
  92         return ret;
  93 }
  94
  95 static int
  96 tcf_ipt_init(struct rtattr *rta, struct rtattr *est, struct tc_action *a, int ovr, int bind)
  97 {
  98         struct ipt_entry_target *t;
  99         unsigned h;
 100         struct rtattr *tb[TCA_IPT_MAX];
 101         struct tcf_ipt *p;
 102         int ret = 0;
 103         u32 index = 0;
 104         u32 hook = 0;
 105
 106         if (NULL == a || NULL == rta ||
 107             (rtattr_parse(tb, TCA_IPT_MAX, RTA_DATA(rta), RTA_PAYLOAD(rta)) <
 108              0)) {
 109                 return -1;
 110         }
 111
 112
 113         if (tb[TCA_IPT_INDEX - 1]) {
 114                 index = *(u32 *) RTA_DATA(tb[TCA_IPT_INDEX - 1]);
 115                 DPRINTK("ipt index %d\n", index);
 116         }
 117
 118         if (index && (p = tcf_hash_lookup(index)) != NULL) {
 119                 a->priv = (void *) p;
 120                 spin_lock(&p->lock);
 121                 if (bind) {
 122                         p->bindcnt += 1;
 123                         p->refcnt += 1;
 124                 }
 125                 if (ovr) {
 126                         goto override;
 127                 }
 128                 spin_unlock(&p->lock);
 129                 return ret;
 130         }
 131
 132         if (NULL == tb[TCA_IPT_TARG - 1] || NULL == tb[TCA_IPT_HOOK - 1]) {
 133                 return -1;
 134         }
 135
 136         p = kmalloc(sizeof (*p), GFP_KERNEL);
 137         if (p == NULL)
 138                 return -1;
 139
 140         memset(p, 0, sizeof (*p));
 141         p->refcnt = 1;
 142         ret = 1;
 143         spin_lock_init(&p->lock);
 144         p->stats_lock = &p->lock;
 145         if (bind)
 146                 p->bindcnt = 1;
 147
 148 override:
 149         hook = *(u32 *) RTA_DATA(tb[TCA_IPT_HOOK - 1]);
 150
 151         t = (struct ipt_entry_target *) RTA_DATA(tb[TCA_IPT_TARG - 1]);
 152
 153         p->t = kmalloc(t->u.target_size, GFP_KERNEL);
 154         if (p->t == NULL) {
 155                 if (ovr) {
 156                         printk("ipt policy messed up \n");
 157                         spin_unlock(&p->lock);
 158                         return -1;
 159                 }
 160                 kfree(p);
 161                 return -1;
 162         }
 163
 164         memcpy(p->t, RTA_DATA(tb[TCA_IPT_TARG - 1]), t->u.target_size);
 165         DPRINTK(" target NAME %s size %d data[0] %x data[1] %x\n",
 166                 t->u.user.name, t->u.target_size, t->data[0], t->data[1]);
 167
 168         p->tname = kmalloc(IFNAMSIZ, GFP_KERNEL);
 169
 170         if (p->tname == NULL) {
 171                 if (ovr) {
 172                         printk("ipt policy messed up 2 \n");
 173                         spin_unlock(&p->lock);
 174                         return -1;
 175                 }
 176                 kfree(p->t);
 177                 kfree(p);
 178                 return -1;
 179         } else {
 180                 int csize = IFNAMSIZ - 1;
 181
 182                 memset(p->tname, 0, IFNAMSIZ);
 183                 if (tb[TCA_IPT_TABLE - 1]) {
 184                         if (strlen((char *) RTA_DATA(tb[TCA_IPT_TABLE - 1])) <
 185                             csize)
 186                                 csize = strlen(RTA_DATA(tb[TCA_IPT_TABLE - 1]));
 187                         strncpy(p->tname, RTA_DATA(tb[TCA_IPT_TABLE - 1]),
 188                                 csize);
 189                         DPRINTK("table name %s\n", p->tname);
 190                 } else {
 191                         strncpy(p->tname, "mangle", 1 + strlen("mangle"));
 192                 }
 193         }
 194
 195         if (0 > init_targ(p)) {
 196                 if (ovr) {
 197                         printk("ipt policy messed up 2 \n");
 198                         spin_unlock(&p->lock);
 199                         return -1;
 200                 }
 201                 kfree(p->tname);
 202                 kfree(p->t);
 203                 kfree(p);
 204                 return -1;
 205         }
 206
 207         if (ovr) {
 208                 spin_unlock(&p->lock);
 209                 return -1;
 210         }
 211
 212         p->index = index ? : tcf_hash_new_index();
 213
 214         p->tm.lastuse = jiffies;
 215         /*
 216         p->tm.expires = jiffies;
 217         */
 218         p->tm.install = jiffies;
 219 #ifdef CONFIG_NET_ESTIMATOR
 220         if (est)
 221                 gen_new_estimator(&p->bstats, &p->rate_est, p->stats_lock, est);
 222 #endif
 223         h = tcf_hash(p->index);
 224         write_lock_bh(&ipt_lock);
 225         p->next = tcf_ipt_ht[h];
 226         tcf_ipt_ht[h] = p;
 227         write_unlock_bh(&ipt_lock);
 228         a->priv = (void *) p;
 229         return ret;
 230
 231 }
 232
 233 static int
 234 tcf_ipt_cleanup(struct tc_action *a, int bind)
 235 {
 236         struct tcf_ipt *p;
 237         p = PRIV(a,ipt);
 238         if (NULL != p)
 239                 return tcf_hash_release(p, bind);
 240         return 0;
 241 }
 242
 243 static int
 244 tcf_ipt(struct sk_buff **pskb, struct tc_action *a)
 245 {
 246         int ret = 0, result = 0;
 247         struct tcf_ipt *p;
 248         struct sk_buff *skb = *pskb;
 249
 250         p = PRIV(a,ipt);
 251
 252         if (NULL == p || NULL == skb) {
 253                 return -1;
 254         }
 255
 256         spin_lock(&p->lock);
 257
 258         p->tm.lastuse = jiffies;
 259         p->bstats.bytes += skb->len;
 260         p->bstats.packets++;
 261
 262         if (skb_cloned(skb) ) {
 263                 if (pskb_expand_head(skb, 0, 0, GFP_ATOMIC)) {
 264                         return -1;
 265                 }
 266         }
 267         /* yes, we have to worry about both in and out dev
 268          worry later - danger - this API seems to have changed
 269          from earlier kernels */
 270
 271         ret = p->t->u.kernel.target->target(&skb, skb->dev, NULL,
 272                                             p->hook, p->t->data, (void *)NULL);
 273         switch (ret) {
 274         case NF_ACCEPT:
 275                 result = TC_ACT_OK;
 276                 break;
 277         case NF_DROP:
 278                 result = TC_ACT_SHOT;
 279                 p->qstats.drops++;
 280                 break;
 281         case IPT_CONTINUE:
 282                 result = TC_ACT_PIPE;
 283                 break;
 284         default:
 285                 if (net_ratelimit())
 286                         printk("Bogus netfilter code %d assume ACCEPT\n", ret);
 287                 result = TC_POLICE_OK;
 288                 break;
 289         }
 290         spin_unlock(&p->lock);
 291         return result;
 292
 293 }
 294
 295 static int
 296 tcf_ipt_dump(struct sk_buff *skb, struct tc_action *a, int bind, int ref)
 297 {
 298         struct ipt_entry_target *t;
 299         struct tcf_t tm;
 300         struct tc_cnt c;
 301         unsigned char *b = skb->tail;
 302
 303         struct tcf_ipt *p;
 304
 305         p = PRIV(a,ipt);
 306         if (NULL == p) {
 307                 printk("BUG: tcf_ipt_dump called with NULL params\n");
 308                 goto rtattr_failure;
 309         }
 310         /* for simple targets kernel size == user size
 311         ** user name = target name
 312         ** for foolproof you need to not assume this
 313         */
 314
 315         t = kmalloc(p->t->u.user.target_size, GFP_ATOMIC);
 316
 317         if (NULL == t)
 318                 goto rtattr_failure;
 319
 320         c.bindcnt = p->bindcnt - bind;
 321         c.refcnt = p->refcnt - ref;
 322         memcpy(t, p->t, p->t->u.user.target_size);
 323         strcpy(t->u.user.name, p->t->u.kernel.target->name);
 324
 325         DPRINTK("\ttcf_ipt_dump tablename %s length %d\n", p->tname,
 326                 strlen(p->tname));
 327         DPRINTK
 328             ("\tdump target name %s size %d size user %d data[0] %x data[1] %x\n",
 329              p->t->u.kernel.target->name, p->t->u.target_size, p->t->u.user.target_size,
 330              p->t->data[0], p->t->data[1]);
 331         RTA_PUT(skb, TCA_IPT_TARG, p->t->u.user.target_size, t);
 332         RTA_PUT(skb, TCA_IPT_INDEX, 4, &p->index);
 333         RTA_PUT(skb, TCA_IPT_HOOK, 4, &p->hook);
 334         RTA_PUT(skb, TCA_IPT_CNT, sizeof(struct tc_cnt), &c);
 335         RTA_PUT(skb, TCA_IPT_TABLE, IFNAMSIZ, p->tname);
 336         tm.install = jiffies_to_clock_t(jiffies - p->tm.install);
 337         tm.lastuse = jiffies_to_clock_t(jiffies - p->tm.lastuse);
 338         tm.expires = jiffies_to_clock_t(p->tm.expires);
 339         RTA_PUT(skb, TCA_IPT_TM, sizeof (tm), &tm);
 340         return skb->len;
 341
 342       rtattr_failure:
 343         skb_trim(skb, b - skb->data);
 344         return -1;
 345 }
 346
 347 static struct tc_action_ops act_ipt_ops = {
 348         .next           =       NULL,
 349         .kind           =       "ipt",
 350         .type           =       TCA_ACT_IPT,
 351         .capab          =       TCA_CAP_NONE,
 352         .owner          =       THIS_MODULE,
 353         .act            =       tcf_ipt,
 354         .dump           =       tcf_ipt_dump,
 355         .cleanup        =       tcf_ipt_cleanup,
 356         .lookup         =       tcf_hash_search,
 357         .init           =       tcf_ipt_init,
 358         .walk           =       tcf_generic_walker
 359 };
 360
 361 MODULE_AUTHOR("Jamal Hadi Salim(2002-4)");
 362 MODULE_DESCRIPTION("Iptables target actions");
 363 MODULE_LICENSE("GPL");
 364
 365 static int __init
 366 ipt_init_module(void)
 367 {
 368         return tcf_register_action(&act_ipt_ops);
 369 }
 370
 371 static void __exit
 372 ipt_cleanup_module(void)
 373 {
 374         tcf_unregister_action(&act_ipt_ops);
 375 }
 376
 377 module_init(ipt_init_module);
 378 module_exit(ipt_cleanup_module);