Fedora kernel-2.6.17-1.2142_FC4 patched with stable patch-2.6.17.4-vs2.0.2-rc26.diff
[linux-2.6.git] / net / ipv4 / netfilter / ip_tables.c
index 8a54f92..cee3397 100644 (file)
@@ -2,7 +2,7 @@
  * Packet matching code.
  *
  * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
- * Copyright (C) 2000-2004 Netfilter Core Team <coreteam@netfilter.org>
+ * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org>
  *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License version 2 as
  * 19 Jan 2002 Harald Welte <laforge@gnumonks.org>
  *     - increase module usage count as soon as we have rules inside
  *       a table
+ * 08 Oct 2005 Harald Welte <lafore@netfilter.org>
+ *     - Generalize into "x_tables" layer and "{ip,ip6,arp}_tables"
  */
 #include <linux/config.h>
 #include <linux/cache.h>
+#include <linux/capability.h>
 #include <linux/skbuff.h>
 #include <linux/kmod.h>
 #include <linux/vmalloc.h>
 #include <linux/netdevice.h>
 #include <linux/module.h>
-#include <linux/tcp.h>
-#include <linux/udp.h>
 #include <linux/icmp.h>
 #include <net/ip.h>
+#include <net/compat.h>
 #include <asm/uaccess.h>
-#include <asm/semaphore.h>
+#include <linux/mutex.h>
 #include <linux/proc_fs.h>
 #include <linux/err.h>
+#include <linux/cpumask.h>
 
+#include <linux/netfilter/x_tables.h>
 #include <linux/netfilter_ipv4/ip_tables.h>
 
 MODULE_LICENSE("GPL");
@@ -60,15 +64,6 @@ do {                                                         \
 #else
 #define IP_NF_ASSERT(x)
 #endif
-#define SMP_ALIGN(x) (((x) + SMP_CACHE_BYTES-1) & ~(SMP_CACHE_BYTES-1))
-
-static DECLARE_MUTEX(ipt_mutex);
-
-/* Must have mutex */
-#define ASSERT_READ_LOCK(x) IP_NF_ASSERT(down_trylock(&ipt_mutex) != 0)
-#define ASSERT_WRITE_LOCK(x) IP_NF_ASSERT(down_trylock(&ipt_mutex) != 0)
-#include <linux/netfilter_ipv4/lockhelp.h>
-#include <linux/netfilter_ipv4/listhelp.h>
 
 #if 0
 /* All the better to debug you with... */
@@ -83,48 +78,8 @@ static DECLARE_MUTEX(ipt_mutex);
    context stops packets coming through and allows user context to read
    the counters or update the rules.
 
-   To be cache friendly on SMP, we arrange them like so:
-   [ n-entries ]
-   ... cache-align padding ...
-   [ n-entries ]
-
    Hence the start of any table is given by get_table() below.  */
 
-/* The table itself */
-struct ipt_table_info
-{
-       /* Size per table */
-       unsigned int size;
-       /* Number of entries: FIXME. --RR */
-       unsigned int number;
-       /* Initial number of entries. Needed for module usage count */
-       unsigned int initial_entries;
-
-       /* Entry points and underflows */
-       unsigned int hook_entry[NF_IP_NUMHOOKS];
-       unsigned int underflow[NF_IP_NUMHOOKS];
-
-       /* ipt_entry tables: one per CPU */
-       char entries[0] ____cacheline_aligned;
-};
-
-static LIST_HEAD(ipt_target);
-static LIST_HEAD(ipt_match);
-static LIST_HEAD(ipt_tables);
-#define ADD_COUNTER(c,b,p) do { (c).bcnt += (b); (c).pcnt += (p); } while(0)
-
-#ifdef CONFIG_SMP
-#define TABLE_OFFSET(t,p) (SMP_ALIGN((t)->size)*(p))
-#else
-#define TABLE_OFFSET(t,p) 0
-#endif
-
-#if 0
-#define down(x) do { printk("DOWN:%u:" #x "\n", __LINE__); down(x); } while(0)
-#define down_interruptible(x) ({ int __r; printk("DOWNi:%u:" #x "\n", __LINE__); __r = down_interruptible(x); if (__r != 0) printk("ABORT-DOWNi:%u\n", __LINE__); __r; })
-#define up(x) do { printk("UP:%u:" #x "\n", __LINE__); up(x); } while(0)
-#endif
-
 /* Returns whether matches rule or not. */
 static inline int
 ip_packet_match(const struct iphdr *ip,
@@ -225,6 +180,7 @@ ipt_error(struct sk_buff **pskb,
          const struct net_device *in,
          const struct net_device *out,
          unsigned int hooknum,
+         const struct xt_target *target,
          const void *targinfo,
          void *userinfo)
 {
@@ -243,7 +199,8 @@ int do_match(struct ipt_entry_match *m,
             int *hotdrop)
 {
        /* Stop iteration if it doesn't match */
-       if (!m->u.kernel.match->match(skb, in, out, m->data, offset, hotdrop))
+       if (!m->u.kernel.match->match(skb, in, out, m->u.kernel.match, m->data,
+                                     offset, skb->nh.iph->ihl*4, hotdrop))
                return 1;
        else
                return 0;
@@ -274,6 +231,7 @@ ipt_do_table(struct sk_buff **pskb,
        const char *indev, *outdev;
        void *table_base;
        struct ipt_entry *e, *back;
+       struct xt_table_info *private = table->private;
 
        /* Initialization */
        ip = (*pskb)->nh.iph;
@@ -290,30 +248,15 @@ ipt_do_table(struct sk_buff **pskb,
 
        read_lock_bh(&table->lock);
        IP_NF_ASSERT(table->valid_hooks & (1 << hook));
-       table_base = (void *)table->private->entries
-               + TABLE_OFFSET(table->private, smp_processor_id());
-       e = get_entry(table_base, table->private->hook_entry[hook]);
-
-#ifdef CONFIG_NETFILTER_DEBUG
-       /* Check noone else using our table */
-       if (((struct ipt_entry *)table_base)->comefrom != 0xdead57ac
-           && ((struct ipt_entry *)table_base)->comefrom != 0xeeeeeeec) {
-               printk("ASSERT: CPU #%u, %s comefrom(%p) = %X\n",
-                      smp_processor_id(),
-                      table->name,
-                      &((struct ipt_entry *)table_base)->comefrom,
-                      ((struct ipt_entry *)table_base)->comefrom);
-       }
-       ((struct ipt_entry *)table_base)->comefrom = 0x57acc001;
-#endif
+       table_base = (void *)private->entries[smp_processor_id()];
+       e = get_entry(table_base, private->hook_entry[hook]);
 
        /* For return from builtin chain */
-       back = get_entry(table_base, table->private->underflow[hook]);
+       back = get_entry(table_base, private->underflow[hook]);
 
        do {
                IP_NF_ASSERT(e);
                IP_NF_ASSERT(back);
-               (*pskb)->nfcache |= e->nfcache;
                if (ip_packet_match(ip, indev, outdev, &e->ip, offset)) {
                        struct ipt_entry_target *t;
 
@@ -342,8 +285,8 @@ ipt_do_table(struct sk_buff **pskb,
                                                         back->comefrom);
                                        continue;
                                }
-                               if (table_base + v
-                                   != (void *)e + e->next_offset) {
+                               if (table_base + v != (void *)e + e->next_offset
+                                   && !(e->ip.flags & IPT_F_GOTO)) {
                                        /* Save old back ptr in next entry */
                                        struct ipt_entry *next
                                                = (void *)e + e->next_offset;
@@ -364,6 +307,7 @@ ipt_do_table(struct sk_buff **pskb,
                                verdict = t->u.kernel.target->target(pskb,
                                                                     in, out,
                                                                     hook,
+                                                                    t->u.kernel.target,
                                                                     t->data,
                                                                     userdata);
 
@@ -395,9 +339,6 @@ ipt_do_table(struct sk_buff **pskb,
                }
        } while (!hotdrop);
 
-#ifdef CONFIG_NETFILTER_DEBUG
-       ((struct ipt_entry *)table_base)->comefrom = 0xdead57ac;
-#endif
        read_unlock_bh(&table->lock);
 
 #ifdef DEBUG_ALLOW_ALL
@@ -409,145 +350,6 @@ ipt_do_table(struct sk_buff **pskb,
 #endif
 }
 
-/*
- * These are weird, but module loading must not be done with mutex
- * held (since they will register), and we have to have a single
- * function to use try_then_request_module().
- */
-
-/* Find table by name, grabs mutex & ref.  Returns ERR_PTR() on error. */
-static inline struct ipt_table *find_table_lock(const char *name)
-{
-       struct ipt_table *t;
-
-       if (down_interruptible(&ipt_mutex) != 0)
-               return ERR_PTR(-EINTR);
-
-       list_for_each_entry(t, &ipt_tables, list)
-               if (strcmp(t->name, name) == 0 && try_module_get(t->me))
-                       return t;
-       up(&ipt_mutex);
-       return NULL;
-}
-
-/* Find match, grabs ref.  Returns ERR_PTR() on error. */
-static inline struct ipt_match *find_match(const char *name, u8 revision)
-{
-       struct ipt_match *m;
-       int err = 0;
-
-       if (down_interruptible(&ipt_mutex) != 0)
-               return ERR_PTR(-EINTR);
-
-       list_for_each_entry(m, &ipt_match, list) {
-               if (strcmp(m->name, name) == 0) {
-                       if (m->revision == revision) {
-                               if (try_module_get(m->me)) {
-                                       up(&ipt_mutex);
-                                       return m;
-                               }
-                       } else
-                               err = -EPROTOTYPE; /* Found something. */
-               }
-       }
-       up(&ipt_mutex);
-       return ERR_PTR(err);
-}
-
-/* Find target, grabs ref.  Returns ERR_PTR() on error. */
-static inline struct ipt_target *find_target(const char *name, u8 revision)
-{
-       struct ipt_target *t;
-       int err = 0;
-
-       if (down_interruptible(&ipt_mutex) != 0)
-               return ERR_PTR(-EINTR);
-
-       list_for_each_entry(t, &ipt_target, list) {
-               if (strcmp(t->name, name) == 0) {
-                       if (t->revision == revision) {
-                               if (try_module_get(t->me)) {
-                                       up(&ipt_mutex);
-                                       return t;
-                               }
-                       } else
-                               err = -EPROTOTYPE; /* Found something. */
-               }
-       }
-       up(&ipt_mutex);
-       return ERR_PTR(err);
-}
-
-struct ipt_target *ipt_find_target(const char *name, u8 revision)
-{
-       struct ipt_target *target;
-
-       target = try_then_request_module(find_target(name, revision),
-                                        "ipt_%s", name);
-       if (IS_ERR(target) || !target)
-               return NULL;
-       return target;
-}
-
-static int match_revfn(const char *name, u8 revision, int *bestp)
-{
-       struct ipt_match *m;
-       int have_rev = 0;
-
-       list_for_each_entry(m, &ipt_match, list) {
-               if (strcmp(m->name, name) == 0) {
-                       if (m->revision > *bestp)
-                               *bestp = m->revision;
-                       if (m->revision == revision)
-                               have_rev = 1;
-               }
-       }
-       return have_rev;
-}
-
-static int target_revfn(const char *name, u8 revision, int *bestp)
-{
-       struct ipt_target *t;
-       int have_rev = 0;
-
-       list_for_each_entry(t, &ipt_target, list) {
-               if (strcmp(t->name, name) == 0) {
-                       if (t->revision > *bestp)
-                               *bestp = t->revision;
-                       if (t->revision == revision)
-                               have_rev = 1;
-               }
-       }
-       return have_rev;
-}
-
-/* Returns true or false (if no such extension at all) */
-static inline int find_revision(const char *name, u8 revision,
-                               int (*revfn)(const char *, u8, int *),
-                               int *err)
-{
-       int have_rev, best = -1;
-
-       if (down_interruptible(&ipt_mutex) != 0) {
-               *err = -EINTR;
-               return 1;
-       }
-       have_rev = revfn(name, revision, &best);
-       up(&ipt_mutex);
-
-       /* Nothing at all?  Return 0 to try loading module. */
-       if (best == -1) {
-               *err = -ENOENT;
-               return 0;
-       }
-
-       *err = best;
-       if (!have_rev)
-               *err = -EPROTONOSUPPORT;
-       return 1;
-}
-
-
 /* All zeroes == unconditional rule. */
 static inline int
 unconditional(const struct ipt_ip *ip)
@@ -564,7 +366,8 @@ unconditional(const struct ipt_ip *ip)
 /* Figures out from what hook each rule can be called: returns 0 if
    there are loops.  Puts hook bitmask in comefrom. */
 static int
-mark_source_chains(struct ipt_table_info *newinfo, unsigned int valid_hooks)
+mark_source_chains(struct xt_table_info *newinfo,
+                  unsigned int valid_hooks, void *entry0)
 {
        unsigned int hook;
 
@@ -573,7 +376,7 @@ mark_source_chains(struct ipt_table_info *newinfo, unsigned int valid_hooks)
        for (hook = 0; hook < NF_IP_NUMHOOKS; hook++) {
                unsigned int pos = newinfo->hook_entry[hook];
                struct ipt_entry *e
-                       = (struct ipt_entry *)(newinfo->entries + pos);
+                       = (struct ipt_entry *)(entry0 + pos);
 
                if (!(valid_hooks & (1 << hook)))
                        continue;
@@ -623,13 +426,13 @@ mark_source_chains(struct ipt_table_info *newinfo, unsigned int valid_hooks)
                                                goto next;
 
                                        e = (struct ipt_entry *)
-                                               (newinfo->entries + pos);
+                                               (entry0 + pos);
                                } while (oldpos == pos + e->next_offset);
 
                                /* Move along one */
                                size = e->next_offset;
                                e = (struct ipt_entry *)
-                                       (newinfo->entries + pos + size);
+                                       (entry0 + pos + size);
                                e->counters.pcnt = pos;
                                pos += size;
                        } else {
@@ -646,7 +449,7 @@ mark_source_chains(struct ipt_table_info *newinfo, unsigned int valid_hooks)
                                        newpos = pos + e->next_offset;
                                }
                                e = (struct ipt_entry *)
-                                       (newinfo->entries + newpos);
+                                       (entry0 + newpos);
                                e->counters.pcnt = pos;
                                pos = newpos;
                        }
@@ -664,7 +467,7 @@ cleanup_match(struct ipt_entry_match *m, unsigned int *i)
                return 1;
 
        if (m->u.kernel.match->destroy)
-               m->u.kernel.match->destroy(m->data,
+               m->u.kernel.match->destroy(m->u.kernel.match, m->data,
                                           m->u.match_size - sizeof(*m));
        module_put(m->u.kernel.match->me);
        return 0;
@@ -677,21 +480,12 @@ standard_check(const struct ipt_entry_target *t,
        struct ipt_standard_target *targ = (void *)t;
 
        /* Check standard info. */
-       if (t->u.target_size
-           != IPT_ALIGN(sizeof(struct ipt_standard_target))) {
-               duprintf("standard_check: target size %u != %u\n",
-                        t->u.target_size,
-                        IPT_ALIGN(sizeof(struct ipt_standard_target)));
-               return 0;
-       }
-
        if (targ->verdict >= 0
            && targ->verdict > max_offset - sizeof(struct ipt_entry)) {
                duprintf("ipt_standard_check: bad verdict (%i)\n",
                         targ->verdict);
                return 0;
        }
-
        if (targ->verdict < -NF_MAX_VERDICT - 1) {
                duprintf("ipt_standard_check: bad negative verdict (%i)\n",
                         targ->verdict);
@@ -708,8 +502,9 @@ check_match(struct ipt_entry_match *m,
            unsigned int *i)
 {
        struct ipt_match *match;
+       int ret;
 
-       match = try_then_request_module(find_match(m->u.user.name,
+       match = try_then_request_module(xt_find_match(AF_INET, m->u.user.name,
                                                   m->u.user.revision),
                                        "ipt_%s", m->u.user.name);
        if (IS_ERR(match) || !match) {
@@ -718,18 +513,27 @@ check_match(struct ipt_entry_match *m,
        }
        m->u.kernel.match = match;
 
+       ret = xt_check_match(match, AF_INET, m->u.match_size - sizeof(*m),
+                            name, hookmask, ip->proto,
+                            ip->invflags & IPT_INV_PROTO);
+       if (ret)
+               goto err;
+
        if (m->u.kernel.match->checkentry
-           && !m->u.kernel.match->checkentry(name, ip, m->data,
+           && !m->u.kernel.match->checkentry(name, ip, match, m->data,
                                              m->u.match_size - sizeof(*m),
                                              hookmask)) {
-               module_put(m->u.kernel.match->me);
                duprintf("ip_tables: check failed for `%s'.\n",
                         m->u.kernel.match->name);
-               return -EINVAL;
+               ret = -EINVAL;
+               goto err;
        }
 
        (*i)++;
        return 0;
+err:
+       module_put(m->u.kernel.match->me);
+       return ret;
 }
 
 static struct ipt_target ipt_standard_target;
@@ -754,7 +558,8 @@ check_entry(struct ipt_entry *e, const char *name, unsigned int size,
                goto cleanup_matches;
 
        t = ipt_get_target(e);
-       target = try_then_request_module(find_target(t->u.user.name,
+       target = try_then_request_module(xt_find_target(AF_INET,
+                                                    t->u.user.name,
                                                     t->u.user.revision),
                                         "ipt_%s", t->u.user.name);
        if (IS_ERR(target) || !target) {
@@ -764,26 +569,32 @@ check_entry(struct ipt_entry *e, const char *name, unsigned int size,
        }
        t->u.kernel.target = target;
 
+       ret = xt_check_target(target, AF_INET, t->u.target_size - sizeof(*t),
+                             name, e->comefrom, e->ip.proto,
+                             e->ip.invflags & IPT_INV_PROTO);
+       if (ret)
+               goto err;
+
        if (t->u.kernel.target == &ipt_standard_target) {
                if (!standard_check(t, size)) {
                        ret = -EINVAL;
                        goto cleanup_matches;
                }
        } else if (t->u.kernel.target->checkentry
-                  && !t->u.kernel.target->checkentry(name, e, t->data,
+                  && !t->u.kernel.target->checkentry(name, e, target, t->data,
                                                      t->u.target_size
                                                      - sizeof(*t),
                                                      e->comefrom)) {
-               module_put(t->u.kernel.target->me);
                duprintf("ip_tables: check failed for `%s'.\n",
                         t->u.kernel.target->name);
                ret = -EINVAL;
-               goto cleanup_matches;
+               goto err;
        }
 
        (*i)++;
        return 0;
-
+ err:
+       module_put(t->u.kernel.target->me);
  cleanup_matches:
        IPT_MATCH_ITERATE(e, cleanup_match, &j);
        return ret;
@@ -791,7 +602,7 @@ check_entry(struct ipt_entry *e, const char *name, unsigned int size,
 
 static inline int
 check_entry_size_and_hooks(struct ipt_entry *e,
-                          struct ipt_table_info *newinfo,
+                          struct xt_table_info *newinfo,
                           unsigned char *base,
                           unsigned char *limit,
                           const unsigned int *hook_entries,
@@ -825,7 +636,7 @@ check_entry_size_and_hooks(struct ipt_entry *e,
            < 0 (not IPT_RETURN). --RR */
 
        /* Clear counters and comefrom */
-       e->counters = ((struct ipt_counters) { 0, 0 });
+       e->counters = ((struct xt_counters) { 0, 0 });
        e->comefrom = 0;
 
        (*i)++;
@@ -844,7 +655,7 @@ cleanup_entry(struct ipt_entry *e, unsigned int *i)
        IPT_MATCH_ITERATE(e, cleanup_match, NULL);
        t = ipt_get_target(e);
        if (t->u.kernel.target->destroy)
-               t->u.kernel.target->destroy(t->data,
+               t->u.kernel.target->destroy(t->u.kernel.target, t->data,
                                            t->u.target_size - sizeof(*t));
        module_put(t->u.kernel.target->me);
        return 0;
@@ -855,7 +666,8 @@ cleanup_entry(struct ipt_entry *e, unsigned int *i)
 static int
 translate_table(const char *name,
                unsigned int valid_hooks,
-               struct ipt_table_info *newinfo,
+               struct xt_table_info *newinfo,
+               void *entry0,
                unsigned int size,
                unsigned int number,
                const unsigned int *hook_entries,
@@ -876,11 +688,11 @@ translate_table(const char *name,
        duprintf("translate_table: size %u\n", newinfo->size);
        i = 0;
        /* Walk through entries, checking offsets. */
-       ret = IPT_ENTRY_ITERATE(newinfo->entries, newinfo->size,
+       ret = IPT_ENTRY_ITERATE(entry0, newinfo->size,
                                check_entry_size_and_hooks,
                                newinfo,
-                               newinfo->entries,
-                               newinfo->entries + size,
+                               entry0,
+                               entry0 + size,
                                hook_entries, underflows, &i);
        if (ret != 0)
                return ret;
@@ -908,93 +720,79 @@ translate_table(const char *name,
                }
        }
 
-       if (!mark_source_chains(newinfo, valid_hooks))
+       if (!mark_source_chains(newinfo, valid_hooks, entry0))
                return -ELOOP;
 
        /* Finally, each sanity check must pass */
        i = 0;
-       ret = IPT_ENTRY_ITERATE(newinfo->entries, newinfo->size,
+       ret = IPT_ENTRY_ITERATE(entry0, newinfo->size,
                                check_entry, name, size, &i);
 
        if (ret != 0) {
-               IPT_ENTRY_ITERATE(newinfo->entries, newinfo->size,
+               IPT_ENTRY_ITERATE(entry0, newinfo->size,
                                  cleanup_entry, &i);
                return ret;
        }
 
        /* And one copy for every other CPU */
-       for (i = 1; i < num_possible_cpus(); i++) {
-               memcpy(newinfo->entries + SMP_ALIGN(newinfo->size)*i,
-                      newinfo->entries,
-                      SMP_ALIGN(newinfo->size));
+       for_each_possible_cpu(i) {
+               if (newinfo->entries[i] && newinfo->entries[i] != entry0)
+                       memcpy(newinfo->entries[i], entry0, newinfo->size);
        }
 
        return ret;
 }
 
-static struct ipt_table_info *
-replace_table(struct ipt_table *table,
-             unsigned int num_counters,
-             struct ipt_table_info *newinfo,
-             int *error)
+/* Gets counters. */
+static inline int
+add_entry_to_counter(const struct ipt_entry *e,
+                    struct xt_counters total[],
+                    unsigned int *i)
 {
-       struct ipt_table_info *oldinfo;
-
-#ifdef CONFIG_NETFILTER_DEBUG
-       {
-               struct ipt_entry *table_base;
-               unsigned int i;
-
-               for (i = 0; i < num_possible_cpus(); i++) {
-                       table_base =
-                               (void *)newinfo->entries
-                               + TABLE_OFFSET(newinfo, i);
-
-                       table_base->comefrom = 0xdead57ac;
-               }
-       }
-#endif
-
-       /* Do the substitution. */
-       write_lock_bh(&table->lock);
-       /* Check inside lock: is the old number correct? */
-       if (num_counters != table->private->number) {
-               duprintf("num_counters != table->private->number (%u/%u)\n",
-                        num_counters, table->private->number);
-               write_unlock_bh(&table->lock);
-               *error = -EAGAIN;
-               return NULL;
-       }
-       oldinfo = table->private;
-       table->private = newinfo;
-       newinfo->initial_entries = oldinfo->initial_entries;
-       write_unlock_bh(&table->lock);
+       ADD_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
 
-       return oldinfo;
+       (*i)++;
+       return 0;
 }
 
-/* Gets counters. */
 static inline int
-add_entry_to_counter(const struct ipt_entry *e,
+set_entry_to_counter(const struct ipt_entry *e,
                     struct ipt_counters total[],
                     unsigned int *i)
 {
-       ADD_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
+       SET_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
 
        (*i)++;
        return 0;
 }
 
 static void
-get_counters(const struct ipt_table_info *t,
-            struct ipt_counters counters[])
+get_counters(const struct xt_table_info *t,
+            struct xt_counters counters[])
 {
        unsigned int cpu;
        unsigned int i;
+       unsigned int curcpu;
+
+       /* Instead of clearing (by a previous call to memset())
+        * the counters and using adds, we set the counters
+        * with data used by 'current' CPU
+        * We dont care about preemption here.
+        */
+       curcpu = raw_smp_processor_id();
+
+       i = 0;
+       IPT_ENTRY_ITERATE(t->entries[curcpu],
+                         t->size,
+                         set_entry_to_counter,
+                         counters,
+                         &i);
 
-       for (cpu = 0; cpu < num_possible_cpus(); cpu++) {
+       for_each_possible_cpu(cpu) {
+               if (cpu == curcpu)
+                       continue;
                i = 0;
-               IPT_ENTRY_ITERATE(t->entries + TABLE_OFFSET(t, cpu),
+               IPT_ENTRY_ITERATE(t->entries[cpu],
                                  t->size,
                                  add_entry_to_counter,
                                  counters,
@@ -1002,33 +800,52 @@ get_counters(const struct ipt_table_info *t,
        }
 }
 
-static int
-copy_entries_to_user(unsigned int total_size,
-                    struct ipt_table *table,
-                    void __user *userptr)
+static inline struct xt_counters * alloc_counters(struct ipt_table *table)
 {
-       unsigned int off, num, countersize;
-       struct ipt_entry *e;
-       struct ipt_counters *counters;
-       int ret = 0;
+       unsigned int countersize;
+       struct xt_counters *counters;
+       struct xt_table_info *private = table->private;
 
        /* We need atomic snapshot of counters: rest doesn't change
           (other than comefrom, which userspace doesn't care
           about). */
-       countersize = sizeof(struct ipt_counters) * table->private->number;
-       counters = vmalloc(countersize);
+       countersize = sizeof(struct xt_counters) * private->number;
+       counters = vmalloc_node(countersize, numa_node_id());
 
        if (counters == NULL)
-               return -ENOMEM;
+               return ERR_PTR(-ENOMEM);
 
        /* First, sum counters... */
-       memset(counters, 0, countersize);
        write_lock_bh(&table->lock);
-       get_counters(table->private, counters);
+       get_counters(private, counters);
        write_unlock_bh(&table->lock);
 
-       /* ... then copy entire thing from CPU 0... */
-       if (copy_to_user(userptr, table->private->entries, total_size) != 0) {
+       return counters;
+}
+
+static int
+copy_entries_to_user(unsigned int total_size,
+                    struct ipt_table *table,
+                    void __user *userptr)
+{
+       unsigned int off, num;
+       struct ipt_entry *e;
+       struct xt_counters *counters;
+       struct xt_table_info *private = table->private;
+       int ret = 0;
+       void *loc_cpu_entry;
+
+       counters = alloc_counters(table);
+       if (IS_ERR(counters))
+               return PTR_ERR(counters);
+
+       /* choose the copy that is on our node/cpu, ...
+        * This choice is lazy (because current thread is
+        * allowed to migrate to another cpu)
+        */
+       loc_cpu_entry = private->entries[raw_smp_processor_id()];
+       /* ... then copy entire thing ... */
+       if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
                ret = -EFAULT;
                goto free_counters;
        }
@@ -1040,7 +857,7 @@ copy_entries_to_user(unsigned int total_size,
                struct ipt_entry_match *m;
                struct ipt_entry_target *t;
 
-               e = (struct ipt_entry *)(table->private->entries + off);
+               e = (struct ipt_entry *)(loc_cpu_entry + off);
                if (copy_to_user(userptr + off
                                 + offsetof(struct ipt_entry, counters),
                                 &counters[num],
@@ -1081,28 +898,277 @@ copy_entries_to_user(unsigned int total_size,
        return ret;
 }
 
+#ifdef CONFIG_COMPAT
+struct compat_delta {
+       struct compat_delta *next;
+       u_int16_t offset;
+       short delta;
+};
+
+static struct compat_delta *compat_offsets = NULL;
+
+static int compat_add_offset(u_int16_t offset, short delta)
+{
+       struct compat_delta *tmp;
+
+       tmp = kmalloc(sizeof(struct compat_delta), GFP_KERNEL);
+       if (!tmp)
+               return -ENOMEM;
+       tmp->offset = offset;
+       tmp->delta = delta;
+       if (compat_offsets) {
+               tmp->next = compat_offsets->next;
+               compat_offsets->next = tmp;
+       } else {
+               compat_offsets = tmp;
+               tmp->next = NULL;
+       }
+       return 0;
+}
+
+static void compat_flush_offsets(void)
+{
+       struct compat_delta *tmp, *next;
+
+       if (compat_offsets) {
+               for(tmp = compat_offsets; tmp; tmp = next) {
+                       next = tmp->next;
+                       kfree(tmp);
+               }
+               compat_offsets = NULL;
+       }
+}
+
+static short compat_calc_jump(u_int16_t offset)
+{
+       struct compat_delta *tmp;
+       short delta;
+
+       for(tmp = compat_offsets, delta = 0; tmp; tmp = tmp->next)
+               if (tmp->offset < offset)
+                       delta += tmp->delta;
+       return delta;
+}
+
+struct compat_ipt_standard_target
+{
+       struct compat_xt_entry_target target;
+       compat_int_t verdict;
+};
+
+struct compat_ipt_standard
+{
+       struct compat_ipt_entry entry;
+       struct compat_ipt_standard_target target;
+};
+
+#define IPT_ST_LEN             XT_ALIGN(sizeof(struct ipt_standard_target))
+#define IPT_ST_COMPAT_LEN      COMPAT_XT_ALIGN(sizeof(struct compat_ipt_standard_target))
+#define IPT_ST_OFFSET          (IPT_ST_LEN - IPT_ST_COMPAT_LEN)
+
+static int compat_ipt_standard_fn(void *target,
+               void **dstptr, int *size, int convert)
+{
+       struct compat_ipt_standard_target compat_st, *pcompat_st;
+       struct ipt_standard_target st, *pst;
+       int ret;
+
+       ret = 0;
+       switch (convert) {
+               case COMPAT_TO_USER:
+                       pst = target;
+                       memcpy(&compat_st.target, &pst->target,
+                               sizeof(compat_st.target));
+                       compat_st.verdict = pst->verdict;
+                       if (compat_st.verdict > 0)
+                               compat_st.verdict -=
+                                       compat_calc_jump(compat_st.verdict);
+                       compat_st.target.u.user.target_size = IPT_ST_COMPAT_LEN;
+                       if (copy_to_user(*dstptr, &compat_st, IPT_ST_COMPAT_LEN))
+                               ret = -EFAULT;
+                       *size -= IPT_ST_OFFSET;
+                       *dstptr += IPT_ST_COMPAT_LEN;
+                       break;
+               case COMPAT_FROM_USER:
+                       pcompat_st = target;
+                       memcpy(&st.target, &pcompat_st->target, IPT_ST_COMPAT_LEN);
+                       st.verdict = pcompat_st->verdict;
+                       if (st.verdict > 0)
+                               st.verdict += compat_calc_jump(st.verdict);
+                       st.target.u.user.target_size = IPT_ST_LEN;
+                       memcpy(*dstptr, &st, IPT_ST_LEN);
+                       *size += IPT_ST_OFFSET;
+                       *dstptr += IPT_ST_LEN;
+                       break;
+               case COMPAT_CALC_SIZE:
+                       *size += IPT_ST_OFFSET;
+                       break;
+               default:
+                       ret = -ENOPROTOOPT;
+                       break;
+       }
+       return ret;
+}
+
+static inline int
+compat_calc_match(struct ipt_entry_match *m, int * size)
+{
+       if (m->u.kernel.match->compat)
+               m->u.kernel.match->compat(m, NULL, size, COMPAT_CALC_SIZE);
+       else
+               xt_compat_match(m, NULL, size, COMPAT_CALC_SIZE);
+       return 0;
+}
+
+static int compat_calc_entry(struct ipt_entry *e, struct xt_table_info *info,
+               void *base, struct xt_table_info *newinfo)
+{
+       struct ipt_entry_target *t;
+       u_int16_t entry_offset;
+       int off, i, ret;
+
+       off = 0;
+       entry_offset = (void *)e - base;
+       IPT_MATCH_ITERATE(e, compat_calc_match, &off);
+       t = ipt_get_target(e);
+       if (t->u.kernel.target->compat)
+               t->u.kernel.target->compat(t, NULL, &off, COMPAT_CALC_SIZE);
+       else
+               xt_compat_target(t, NULL, &off, COMPAT_CALC_SIZE);
+       newinfo->size -= off;
+       ret = compat_add_offset(entry_offset, off);
+       if (ret)
+               return ret;
+
+       for (i = 0; i< NF_IP_NUMHOOKS; i++) {
+               if (info->hook_entry[i] && (e < (struct ipt_entry *)
+                               (base + info->hook_entry[i])))
+                       newinfo->hook_entry[i] -= off;
+               if (info->underflow[i] && (e < (struct ipt_entry *)
+                               (base + info->underflow[i])))
+                       newinfo->underflow[i] -= off;
+       }
+       return 0;
+}
+
+static int compat_table_info(struct xt_table_info *info,
+               struct xt_table_info *newinfo)
+{
+       void *loc_cpu_entry;
+       int i;
+
+       if (!newinfo || !info)
+               return -EINVAL;
+
+       memset(newinfo, 0, sizeof(struct xt_table_info));
+       newinfo->size = info->size;
+       newinfo->number = info->number;
+       for (i = 0; i < NF_IP_NUMHOOKS; i++) {
+               newinfo->hook_entry[i] = info->hook_entry[i];
+               newinfo->underflow[i] = info->underflow[i];
+       }
+       loc_cpu_entry = info->entries[raw_smp_processor_id()];
+       return IPT_ENTRY_ITERATE(loc_cpu_entry, info->size,
+                       compat_calc_entry, info, loc_cpu_entry, newinfo);
+}
+#endif
+
+static int get_info(void __user *user, int *len, int compat)
+{
+       char name[IPT_TABLE_MAXNAMELEN];
+       struct ipt_table *t;
+       int ret;
+
+       if (*len != sizeof(struct ipt_getinfo)) {
+               duprintf("length %u != %u\n", *len,
+                       (unsigned int)sizeof(struct ipt_getinfo));
+               return -EINVAL;
+       }
+
+       if (copy_from_user(name, user, sizeof(name)) != 0)
+               return -EFAULT;
+
+       name[IPT_TABLE_MAXNAMELEN-1] = '\0';
+#ifdef CONFIG_COMPAT
+       if (compat)
+               xt_compat_lock(AF_INET);
+#endif
+       t = try_then_request_module(xt_find_table_lock(AF_INET, name),
+                       "iptable_%s", name);
+       if (t && !IS_ERR(t)) {
+               struct ipt_getinfo info;
+               struct xt_table_info *private = t->private;
+
+#ifdef CONFIG_COMPAT
+               if (compat) {
+                       struct xt_table_info tmp;
+                       ret = compat_table_info(private, &tmp);
+                       compat_flush_offsets();
+                       private =  &tmp;
+               }
+#endif
+               info.valid_hooks = t->valid_hooks;
+               memcpy(info.hook_entry, private->hook_entry,
+                               sizeof(info.hook_entry));
+               memcpy(info.underflow, private->underflow,
+                               sizeof(info.underflow));
+               info.num_entries = private->number;
+               info.size = private->size;
+               strcpy(info.name, name);
+
+               if (copy_to_user(user, &info, *len) != 0)
+                       ret = -EFAULT;
+               else
+                       ret = 0;
+
+               xt_table_unlock(t);
+               module_put(t->me);
+       } else
+               ret = t ? PTR_ERR(t) : -ENOENT;
+#ifdef CONFIG_COMPAT
+       if (compat)
+               xt_compat_unlock(AF_INET);
+#endif
+       return ret;
+}
+
 static int
-get_entries(const struct ipt_get_entries *entries,
-           struct ipt_get_entries __user *uptr)
+get_entries(struct ipt_get_entries __user *uptr, int *len)
 {
        int ret;
+       struct ipt_get_entries get;
        struct ipt_table *t;
 
-       t = find_table_lock(entries->name);
+       if (*len < sizeof(get)) {
+               duprintf("get_entries: %u < %d\n", *len,
+                               (unsigned int)sizeof(get));
+               return -EINVAL;
+       }
+       if (copy_from_user(&get, uptr, sizeof(get)) != 0)
+               return -EFAULT;
+       if (*len != sizeof(struct ipt_get_entries) + get.size) {
+               duprintf("get_entries: %u != %u\n", *len,
+                               (unsigned int)(sizeof(struct ipt_get_entries) +
+                               get.size));
+               return -EINVAL;
+       }
+
+       t = xt_find_table_lock(AF_INET, get.name);
        if (t && !IS_ERR(t)) {
+               struct xt_table_info *private = t->private;
                duprintf("t->private->number = %u\n",
-                        t->private->number);
-               if (entries->size == t->private->size)
-                       ret = copy_entries_to_user(t->private->size,
+                        private->number);
+               if (get.size == private->size)
+                       ret = copy_entries_to_user(private->size,
                                                   t, uptr->entrytable);
                else {
                        duprintf("get_entries: I've got %u not %u!\n",
-                                t->private->size,
-                                entries->size);
+                                private->size,
+                                get.size);
                        ret = -EINVAL;
                }
                module_put(t->me);
-               up(&ipt_mutex);
+               xt_table_unlock(t);
        } else
                ret = t ? PTR_ERR(t) : -ENOENT;
 
@@ -1110,75 +1176,47 @@ get_entries(const struct ipt_get_entries *entries,
 }
 
 static int
-do_replace(void __user *user, unsigned int len)
+__do_replace(const char *name, unsigned int valid_hooks,
+               struct xt_table_info *newinfo, unsigned int num_counters,
+               void __user *counters_ptr)
 {
        int ret;
-       struct ipt_replace tmp;
        struct ipt_table *t;
-       struct ipt_table_info *newinfo, *oldinfo;
-       struct ipt_counters *counters;
-
-       if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
-               return -EFAULT;
-
-       /* Hack: Causes ipchains to give correct error msg --RR */
-       if (len != sizeof(tmp) + tmp.size)
-               return -ENOPROTOOPT;
-
-       /* Pedantry: prevent them from hitting BUG() in vmalloc.c --RR */
-       if ((SMP_ALIGN(tmp.size) >> PAGE_SHIFT) + 2 > num_physpages)
-               return -ENOMEM;
-
-       newinfo = vmalloc(sizeof(struct ipt_table_info)
-                         + SMP_ALIGN(tmp.size) * num_possible_cpus());
-       if (!newinfo)
-               return -ENOMEM;
-
-       if (copy_from_user(newinfo->entries, user + sizeof(tmp),
-                          tmp.size) != 0) {
-               ret = -EFAULT;
-               goto free_newinfo;
-       }
+       struct xt_table_info *oldinfo;
+       struct xt_counters *counters;
+       void *loc_cpu_old_entry;
 
-       counters = vmalloc(tmp.num_counters * sizeof(struct ipt_counters));
+       ret = 0;
+       counters = vmalloc(num_counters * sizeof(struct xt_counters));
        if (!counters) {
                ret = -ENOMEM;
-               goto free_newinfo;
+               goto out;
        }
-       memset(counters, 0, tmp.num_counters * sizeof(struct ipt_counters));
-
-       ret = translate_table(tmp.name, tmp.valid_hooks,
-                             newinfo, tmp.size, tmp.num_entries,
-                             tmp.hook_entry, tmp.underflow);
-       if (ret != 0)
-               goto free_newinfo_counters;
-
-       duprintf("ip_tables: Translated table\n");
 
-       t = try_then_request_module(find_table_lock(tmp.name),
-                                   "iptable_%s", tmp.name);
+       t = try_then_request_module(xt_find_table_lock(AF_INET, name),
+                                   "iptable_%s", name);
        if (!t || IS_ERR(t)) {
                ret = t ? PTR_ERR(t) : -ENOENT;
                goto free_newinfo_counters_untrans;
        }
 
        /* You lied! */
-       if (tmp.valid_hooks != t->valid_hooks) {
+       if (valid_hooks != t->valid_hooks) {
                duprintf("Valid hook crap: %08X vs %08X\n",
-                        tmp.valid_hooks, t->valid_hooks);
+                        valid_hooks, t->valid_hooks);
                ret = -EINVAL;
                goto put_module;
        }
 
-       oldinfo = replace_table(t, tmp.num_counters, newinfo, &ret);
+       oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
        if (!oldinfo)
                goto put_module;
 
        /* Update module usage count based on number of rules */
        duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
                oldinfo->number, oldinfo->initial_entries, newinfo->number);
-       if ((oldinfo->number > oldinfo->initial_entries) || 
-           (newinfo->number <= oldinfo->initial_entries)) 
+       if ((oldinfo->number > oldinfo->initial_entries) ||
+           (newinfo->number <= oldinfo->initial_entries))
                module_put(t->me);
        if ((oldinfo->number > oldinfo->initial_entries) &&
            (newinfo->number <= oldinfo->initial_entries))
@@ -1187,33 +1225,87 @@ do_replace(void __user *user, unsigned int len)
        /* Get the old counters. */
        get_counters(oldinfo, counters);
        /* Decrease module usage counts and free resource */
-       IPT_ENTRY_ITERATE(oldinfo->entries, oldinfo->size, cleanup_entry,NULL);
-       vfree(oldinfo);
-       if (copy_to_user(tmp.counters, counters,
-                        sizeof(struct ipt_counters) * tmp.num_counters) != 0)
+       loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
+       IPT_ENTRY_ITERATE(loc_cpu_old_entry, oldinfo->size, cleanup_entry,NULL);
+       xt_free_table_info(oldinfo);
+       if (copy_to_user(counters_ptr, counters,
+                        sizeof(struct xt_counters) * num_counters) != 0)
                ret = -EFAULT;
        vfree(counters);
-       up(&ipt_mutex);
+       xt_table_unlock(t);
        return ret;
 
  put_module:
        module_put(t->me);
-       up(&ipt_mutex);
+       xt_table_unlock(t);
  free_newinfo_counters_untrans:
-       IPT_ENTRY_ITERATE(newinfo->entries, newinfo->size, cleanup_entry,NULL);
- free_newinfo_counters:
        vfree(counters);
- free_newinfo:
-       vfree(newinfo);
+ out:
        return ret;
 }
 
-/* We're lazy, and add to the first CPU; overflow works its fey magic
- * and everything is OK. */
-static inline int
-add_counter_to_entry(struct ipt_entry *e,
-                    const struct ipt_counters addme[],
-                    unsigned int *i)
+static int
+do_replace(void __user *user, unsigned int len)
+{
+       int ret;
+       struct ipt_replace tmp;
+       struct xt_table_info *newinfo;
+       void *loc_cpu_entry;
+
+       if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
+               return -EFAULT;
+
+       /* Hack: Causes ipchains to give correct error msg --RR */
+       if (len != sizeof(tmp) + tmp.size)
+               return -ENOPROTOOPT;
+
+       /* overflow check */
+       if (tmp.size >= (INT_MAX - sizeof(struct xt_table_info)) / NR_CPUS -
+                       SMP_CACHE_BYTES)
+               return -ENOMEM;
+       if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
+               return -ENOMEM;
+
+       newinfo = xt_alloc_table_info(tmp.size);
+       if (!newinfo)
+               return -ENOMEM;
+
+       /* choose the copy that is our node/cpu */
+       loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
+       if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
+                          tmp.size) != 0) {
+               ret = -EFAULT;
+               goto free_newinfo;
+       }
+
+       ret = translate_table(tmp.name, tmp.valid_hooks,
+                             newinfo, loc_cpu_entry, tmp.size, tmp.num_entries,
+                             tmp.hook_entry, tmp.underflow);
+       if (ret != 0)
+               goto free_newinfo;
+
+       duprintf("ip_tables: Translated table\n");
+
+       ret = __do_replace(tmp.name, tmp.valid_hooks,
+                             newinfo, tmp.num_counters,
+                             tmp.counters);
+       if (ret)
+               goto free_newinfo_untrans;
+       return 0;
+
+ free_newinfo_untrans:
+       IPT_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry,NULL);
+ free_newinfo:
+       xt_free_table_info(newinfo);
+       return ret;
+}
+
+/* We're lazy, and add to the first CPU; overflow works its fey magic
+ * and everything is OK. */
+static inline int
+add_counter_to_entry(struct ipt_entry *e,
+                    const struct xt_counters addme[],
+                    unsigned int *i)
 {
 #if 0
        duprintf("add_counter: Entry %u %lu/%lu + %lu/%lu\n",
@@ -1231,49 +1323,82 @@ add_counter_to_entry(struct ipt_entry *e,
 }
 
 static int
-do_add_counters(void __user *user, unsigned int len)
+do_add_counters(void __user *user, unsigned int len, int compat)
 {
        unsigned int i;
-       struct ipt_counters_info tmp, *paddc;
+       struct xt_counters_info tmp;
+       struct xt_counters *paddc;
+       unsigned int num_counters;
+       char *name;
+       int size;
+       void *ptmp;
        struct ipt_table *t;
+       struct xt_table_info *private;
        int ret = 0;
+       void *loc_cpu_entry;
+#ifdef CONFIG_COMPAT
+       struct compat_xt_counters_info compat_tmp;
 
-       if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
+       if (compat) {
+               ptmp = &compat_tmp;
+               size = sizeof(struct compat_xt_counters_info);
+       } else
+#endif
+       {
+               ptmp = &tmp;
+               size = sizeof(struct xt_counters_info);
+       }
+
+       if (copy_from_user(ptmp, user, size) != 0)
                return -EFAULT;
 
-       if (len != sizeof(tmp) + tmp.num_counters*sizeof(struct ipt_counters))
+#ifdef CONFIG_COMPAT
+       if (compat) {
+               num_counters = compat_tmp.num_counters;
+               name = compat_tmp.name;
+       } else
+#endif
+       {
+               num_counters = tmp.num_counters;
+               name = tmp.name;
+       }
+
+       if (len != size + num_counters * sizeof(struct xt_counters))
                return -EINVAL;
 
-       paddc = vmalloc(len);
+       paddc = vmalloc_node(len - size, numa_node_id());
        if (!paddc)
                return -ENOMEM;
 
-       if (copy_from_user(paddc, user, len) != 0) {
+       if (copy_from_user(paddc, user + size, len - size) != 0) {
                ret = -EFAULT;
                goto free;
        }
 
-       t = find_table_lock(tmp.name);
+       t = xt_find_table_lock(AF_INET, name);
        if (!t || IS_ERR(t)) {
                ret = t ? PTR_ERR(t) : -ENOENT;
                goto free;
        }
 
        write_lock_bh(&t->lock);
-       if (t->private->number != paddc->num_counters) {
+       private = t->private;
+       if (private->number != num_counters) {
                ret = -EINVAL;
                goto unlock_up_free;
        }
 
        i = 0;
-       IPT_ENTRY_ITERATE(t->private->entries,
-                         t->private->size,
+       /* Choose the copy that is on our node */
+       loc_cpu_entry = private->entries[raw_smp_processor_id()];
+       IPT_ENTRY_ITERATE(loc_cpu_entry,
+                         private->size,
                          add_counter_to_entry,
-                         paddc->counters,
+                         paddc,
                          &i);
  unlock_up_free:
        write_unlock_bh(&t->lock);
-       up(&ipt_mutex);
+       xt_table_unlock(t);
        module_put(t->me);
  free:
        vfree(paddc);
@@ -1281,434 +1406,732 @@ do_add_counters(void __user *user, unsigned int len)
        return ret;
 }
 
-static int
-do_ipt_set_ctl(struct sock *sk,        int cmd, void __user *user, unsigned int len)
-{
-       int ret;
-
-       if (!capable(CAP_NET_ADMIN))
-               return -EPERM;
-
-       switch (cmd) {
-       case IPT_SO_SET_REPLACE:
-               ret = do_replace(user, len);
-               break;
+#ifdef CONFIG_COMPAT
+struct compat_ipt_replace {
+       char                    name[IPT_TABLE_MAXNAMELEN];
+       u32                     valid_hooks;
+       u32                     num_entries;
+       u32                     size;
+       u32                     hook_entry[NF_IP_NUMHOOKS];
+       u32                     underflow[NF_IP_NUMHOOKS];
+       u32                     num_counters;
+       compat_uptr_t           counters;       /* struct ipt_counters * */
+       struct compat_ipt_entry entries[0];
+};
 
-       case IPT_SO_SET_ADD_COUNTERS:
-               ret = do_add_counters(user, len);
-               break;
+static inline int compat_copy_match_to_user(struct ipt_entry_match *m,
+               void __user **dstptr, compat_uint_t *size)
+{
+       if (m->u.kernel.match->compat)
+               return m->u.kernel.match->compat(m, dstptr, size,
+                               COMPAT_TO_USER);
+       else
+               return xt_compat_match(m, dstptr, size, COMPAT_TO_USER);
+}
 
-       default:
-               duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
-               ret = -EINVAL;
-       }
+static int compat_copy_entry_to_user(struct ipt_entry *e,
+               void __user **dstptr, compat_uint_t *size)
+{
+       struct ipt_entry_target __user *t;
+       struct compat_ipt_entry __user *ce;
+       u_int16_t target_offset, next_offset;
+       compat_uint_t origsize;
+       int ret;
 
+       ret = -EFAULT;
+       origsize = *size;
+       ce = (struct compat_ipt_entry __user *)*dstptr;
+       if (copy_to_user(ce, e, sizeof(struct ipt_entry)))
+               goto out;
+
+       *dstptr += sizeof(struct compat_ipt_entry);
+       ret = IPT_MATCH_ITERATE(e, compat_copy_match_to_user, dstptr, size);
+       target_offset = e->target_offset - (origsize - *size);
+       if (ret)
+               goto out;
+       t = ipt_get_target(e);
+       if (t->u.kernel.target->compat)
+               ret = t->u.kernel.target->compat(t, dstptr, size,
+                               COMPAT_TO_USER);
+       else
+               ret = xt_compat_target(t, dstptr, size, COMPAT_TO_USER);
+       if (ret)
+               goto out;
+       ret = -EFAULT;
+       next_offset = e->next_offset - (origsize - *size);
+       if (put_user(target_offset, &ce->target_offset))
+               goto out;
+       if (put_user(next_offset, &ce->next_offset))
+               goto out;
+       return 0;
+out:
        return ret;
 }
 
-static int
-do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
+static inline int
+compat_check_calc_match(struct ipt_entry_match *m,
+           const char *name,
+           const struct ipt_ip *ip,
+           unsigned int hookmask,
+           int *size, int *i)
 {
-       int ret;
+       struct ipt_match *match;
 
-       if (!capable(CAP_NET_ADMIN))
-               return -EPERM;
+       match = try_then_request_module(xt_find_match(AF_INET, m->u.user.name,
+                                                  m->u.user.revision),
+                                       "ipt_%s", m->u.user.name);
+       if (IS_ERR(match) || !match) {
+               duprintf("compat_check_calc_match: `%s' not found\n",
+                               m->u.user.name);
+               return match ? PTR_ERR(match) : -ENOENT;
+       }
+       m->u.kernel.match = match;
 
-       switch (cmd) {
-       case IPT_SO_GET_INFO: {
-               char name[IPT_TABLE_MAXNAMELEN];
-               struct ipt_table *t;
+       if (m->u.kernel.match->compat)
+               m->u.kernel.match->compat(m, NULL, size, COMPAT_CALC_SIZE);
+       else
+               xt_compat_match(m, NULL, size, COMPAT_CALC_SIZE);
 
-               if (*len != sizeof(struct ipt_getinfo)) {
-                       duprintf("length %u != %u\n", *len,
-                                sizeof(struct ipt_getinfo));
-                       ret = -EINVAL;
-                       break;
-               }
+       (*i)++;
+       return 0;
+}
 
-               if (copy_from_user(name, user, sizeof(name)) != 0) {
-                       ret = -EFAULT;
-                       break;
-               }
-               name[IPT_TABLE_MAXNAMELEN-1] = '\0';
-
-               t = try_then_request_module(find_table_lock(name),
-                                           "iptable_%s", name);
-               if (t && !IS_ERR(t)) {
-                       struct ipt_getinfo info;
-
-                       info.valid_hooks = t->valid_hooks;
-                       memcpy(info.hook_entry, t->private->hook_entry,
-                              sizeof(info.hook_entry));
-                       memcpy(info.underflow, t->private->underflow,
-                              sizeof(info.underflow));
-                       info.num_entries = t->private->number;
-                       info.size = t->private->size;
-                       memcpy(info.name, name, sizeof(info.name));
-
-                       if (copy_to_user(user, &info, *len) != 0)
-                               ret = -EFAULT;
-                       else
-                               ret = 0;
-                       up(&ipt_mutex);
-                       module_put(t->me);
-               } else
-                       ret = t ? PTR_ERR(t) : -ENOENT;
+static inline int
+check_compat_entry_size_and_hooks(struct ipt_entry *e,
+                          struct xt_table_info *newinfo,
+                          unsigned int *size,
+                          unsigned char *base,
+                          unsigned char *limit,
+                          unsigned int *hook_entries,
+                          unsigned int *underflows,
+                          unsigned int *i,
+                          const char *name)
+{
+       struct ipt_entry_target *t;
+       struct ipt_target *target;
+       u_int16_t entry_offset;
+       int ret, off, h, j;
+
+       duprintf("check_compat_entry_size_and_hooks %p\n", e);
+       if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0
+           || (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
+               duprintf("Bad offset %p, limit = %p\n", e, limit);
+               return -EINVAL;
        }
-       break;
 
-       case IPT_SO_GET_ENTRIES: {
-               struct ipt_get_entries get;
+       if (e->next_offset < sizeof(struct compat_ipt_entry) +
+                       sizeof(struct compat_xt_entry_target)) {
+               duprintf("checking: element %p size %u\n",
+                        e, e->next_offset);
+               return -EINVAL;
+       }
 
-               if (*len < sizeof(get)) {
-                       duprintf("get_entries: %u < %u\n", *len, sizeof(get));
-                       ret = -EINVAL;
-               } else if (copy_from_user(&get, user, sizeof(get)) != 0) {
-                       ret = -EFAULT;
-               } else if (*len != sizeof(struct ipt_get_entries) + get.size) {
-                       duprintf("get_entries: %u != %u\n", *len,
-                                sizeof(struct ipt_get_entries) + get.size);
-                       ret = -EINVAL;
-               } else
-                       ret = get_entries(&get, user);
-               break;
+       if (!ip_checkentry(&e->ip)) {
+               duprintf("ip_tables: ip check failed %p %s.\n", e, name);
+               return -EINVAL;
        }
 
-       case IPT_SO_GET_REVISION_MATCH:
-       case IPT_SO_GET_REVISION_TARGET: {
-               struct ipt_get_revision rev;
-               int (*revfn)(const char *, u8, int *);
+       off = 0;
+       entry_offset = (void *)e - (void *)base;
+       j = 0;
+       ret = IPT_MATCH_ITERATE(e, compat_check_calc_match, name, &e->ip,
+                       e->comefrom, &off, &j);
+       if (ret != 0)
+               goto out;
 
-               if (*len != sizeof(rev)) {
-                       ret = -EINVAL;
-                       break;
-               }
-               if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
-                       ret = -EFAULT;
-                       break;
-               }
+       t = ipt_get_target(e);
+       target = try_then_request_module(xt_find_target(AF_INET,
+                                                    t->u.user.name,
+                                                    t->u.user.revision),
+                                        "ipt_%s", t->u.user.name);
+       if (IS_ERR(target) || !target) {
+               duprintf("check_entry: `%s' not found\n", t->u.user.name);
+               ret = target ? PTR_ERR(target) : -ENOENT;
+               goto out;
+       }
+       t->u.kernel.target = target;
 
-               if (cmd == IPT_SO_GET_REVISION_TARGET)
-                       revfn = target_revfn;
-               else
-                       revfn = match_revfn;
+       if (t->u.kernel.target->compat)
+               t->u.kernel.target->compat(t, NULL, &off, COMPAT_CALC_SIZE);
+       else
+               xt_compat_target(t, NULL, &off, COMPAT_CALC_SIZE);
+       *size += off;
+       ret = compat_add_offset(entry_offset, off);
+       if (ret)
+               goto out;
 
-               try_then_request_module(find_revision(rev.name, rev.revision,
-                                                     revfn, &ret),
-                                       "ipt_%s", rev.name);
-               break;
+       /* Check hooks & underflows */
+       for (h = 0; h < NF_IP_NUMHOOKS; h++) {
+               if ((unsigned char *)e - base == hook_entries[h])
+                       newinfo->hook_entry[h] = hook_entries[h];
+               if ((unsigned char *)e - base == underflows[h])
+                       newinfo->underflow[h] = underflows[h];
        }
 
-       default:
-               duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
-               ret = -EINVAL;
-       }
+       /* Clear counters and comefrom */
+       e->counters = ((struct ipt_counters) { 0, 0 });
+       e->comefrom = 0;
 
+       (*i)++;
+       return 0;
+out:
+       IPT_MATCH_ITERATE(e, cleanup_match, &j);
        return ret;
 }
 
-/* Registration hooks for targets. */
-int
-ipt_register_target(struct ipt_target *target)
+static inline int compat_copy_match_from_user(struct ipt_entry_match *m,
+       void **dstptr, compat_uint_t *size, const char *name,
+       const struct ipt_ip *ip, unsigned int hookmask)
 {
+       struct ipt_entry_match *dm;
+       struct ipt_match *match;
        int ret;
 
-       ret = down_interruptible(&ipt_mutex);
-       if (ret != 0)
+       dm = (struct ipt_entry_match *)*dstptr;
+       match = m->u.kernel.match;
+       if (match->compat)
+               match->compat(m, dstptr, size, COMPAT_FROM_USER);
+       else
+               xt_compat_match(m, dstptr, size, COMPAT_FROM_USER);
+
+       ret = xt_check_match(match, AF_INET, dm->u.match_size - sizeof(*dm),
+                            name, hookmask, ip->proto,
+                            ip->invflags & IPT_INV_PROTO);
+       if (ret)
                return ret;
-       list_add(&target->list, &ipt_target);
-       up(&ipt_mutex);
-       return ret;
-}
 
-void
-ipt_unregister_target(struct ipt_target *target)
-{
-       down(&ipt_mutex);
-       LIST_DELETE(&ipt_target, target);
-       up(&ipt_mutex);
+       if (m->u.kernel.match->checkentry
+           && !m->u.kernel.match->checkentry(name, ip, match, dm->data,
+                                             dm->u.match_size - sizeof(*dm),
+                                             hookmask)) {
+               duprintf("ip_tables: check failed for `%s'.\n",
+                        m->u.kernel.match->name);
+               return -EINVAL;
+       }
+       return 0;
 }
 
-int
-ipt_register_match(struct ipt_match *match)
+static int compat_copy_entry_from_user(struct ipt_entry *e, void **dstptr,
+       unsigned int *size, const char *name,
+       struct xt_table_info *newinfo, unsigned char *base)
 {
-       int ret;
+       struct ipt_entry_target *t;
+       struct ipt_target *target;
+       struct ipt_entry *de;
+       unsigned int origsize;
+       int ret, h;
+
+       ret = 0;
+       origsize = *size;
+       de = (struct ipt_entry *)*dstptr;
+       memcpy(de, e, sizeof(struct ipt_entry));
+
+       *dstptr += sizeof(struct compat_ipt_entry);
+       ret = IPT_MATCH_ITERATE(e, compat_copy_match_from_user, dstptr, size,
+                       name, &de->ip, de->comefrom);
+       if (ret)
+               goto out;
+       de->target_offset = e->target_offset - (origsize - *size);
+       t = ipt_get_target(e);
+       target = t->u.kernel.target;
+       if (target->compat)
+               target->compat(t, dstptr, size, COMPAT_FROM_USER);
+       else
+               xt_compat_target(t, dstptr, size, COMPAT_FROM_USER);
 
-       ret = down_interruptible(&ipt_mutex);
-       if (ret != 0)
-               return ret;
+       de->next_offset = e->next_offset - (origsize - *size);
+       for (h = 0; h < NF_IP_NUMHOOKS; h++) {
+               if ((unsigned char *)de - base < newinfo->hook_entry[h])
+                       newinfo->hook_entry[h] -= origsize - *size;
+               if ((unsigned char *)de - base < newinfo->underflow[h])
+                       newinfo->underflow[h] -= origsize - *size;
+       }
 
-       list_add(&match->list, &ipt_match);
-       up(&ipt_mutex);
+       t = ipt_get_target(de);
+       target = t->u.kernel.target;
+       ret = xt_check_target(target, AF_INET, t->u.target_size - sizeof(*t),
+                             name, e->comefrom, e->ip.proto,
+                             e->ip.invflags & IPT_INV_PROTO);
+       if (ret)
+               goto out;
 
+       ret = -EINVAL;
+       if (t->u.kernel.target == &ipt_standard_target) {
+               if (!standard_check(t, *size))
+                       goto out;
+       } else if (t->u.kernel.target->checkentry
+                  && !t->u.kernel.target->checkentry(name, de, target,
+                               t->data, t->u.target_size - sizeof(*t),
+                               de->comefrom)) {
+               duprintf("ip_tables: compat: check failed for `%s'.\n",
+                        t->u.kernel.target->name);
+               goto out;
+       }
+       ret = 0;
+out:
        return ret;
 }
 
-void
-ipt_unregister_match(struct ipt_match *match)
-{
-       down(&ipt_mutex);
-       LIST_DELETE(&ipt_match, match);
-       up(&ipt_mutex);
-}
-
-int ipt_register_table(struct ipt_table *table, const struct ipt_replace *repl)
+static int
+translate_compat_table(const char *name,
+               unsigned int valid_hooks,
+               struct xt_table_info **pinfo,
+               void **pentry0,
+               unsigned int total_size,
+               unsigned int number,
+               unsigned int *hook_entries,
+               unsigned int *underflows)
 {
+       unsigned int i;
+       struct xt_table_info *newinfo, *info;
+       void *pos, *entry0, *entry1;
+       unsigned int size;
        int ret;
-       struct ipt_table_info *newinfo;
-       static struct ipt_table_info bootstrap
-               = { 0, 0, 0, { 0 }, { 0 }, { } };
 
-       newinfo = vmalloc(sizeof(struct ipt_table_info)
-                         + SMP_ALIGN(repl->size) * num_possible_cpus());
-       if (!newinfo)
-               return -ENOMEM;
+       info = *pinfo;
+       entry0 = *pentry0;
+       size = total_size;
+       info->number = number;
 
-       memcpy(newinfo->entries, repl->entries, repl->size);
-
-       ret = translate_table(table->name, table->valid_hooks,
-                             newinfo, repl->size,
-                             repl->num_entries,
-                             repl->hook_entry,
-                             repl->underflow);
-       if (ret != 0) {
-               vfree(newinfo);
-               return ret;
+       /* Init all hooks to impossible value. */
+       for (i = 0; i < NF_IP_NUMHOOKS; i++) {
+               info->hook_entry[i] = 0xFFFFFFFF;
+               info->underflow[i] = 0xFFFFFFFF;
        }
 
-       ret = down_interruptible(&ipt_mutex);
-       if (ret != 0) {
-               vfree(newinfo);
-               return ret;
+       duprintf("translate_compat_table: size %u\n", info->size);
+       i = 0;
+       xt_compat_lock(AF_INET);
+       /* Walk through entries, checking offsets. */
+       ret = IPT_ENTRY_ITERATE(entry0, total_size,
+                               check_compat_entry_size_and_hooks,
+                               info, &size, entry0,
+                               entry0 + total_size,
+                               hook_entries, underflows, &i, name);
+       if (ret != 0)
+               goto out_unlock;
+
+       ret = -EINVAL;
+       if (i != number) {
+               duprintf("translate_compat_table: %u not %u entries\n",
+                        i, number);
+               goto out_unlock;
        }
 
-       /* Don't autoload: we'd eat our tail... */
-       if (list_named_find(&ipt_tables, table->name)) {
-               ret = -EEXIST;
-               goto free_unlock;
+       /* Check hooks all assigned */
+       for (i = 0; i < NF_IP_NUMHOOKS; i++) {
+               /* Only hooks which are valid */
+               if (!(valid_hooks & (1 << i)))
+                       continue;
+               if (info->hook_entry[i] == 0xFFFFFFFF) {
+                       duprintf("Invalid hook entry %u %u\n",
+                                i, hook_entries[i]);
+                       goto out_unlock;
+               }
+               if (info->underflow[i] == 0xFFFFFFFF) {
+                       duprintf("Invalid underflow %u %u\n",
+                                i, underflows[i]);
+                       goto out_unlock;
+               }
        }
 
-       /* Simplifies replace_table code. */
-       table->private = &bootstrap;
-       if (!replace_table(table, 0, newinfo, &ret))
-               goto free_unlock;
+       ret = -ENOMEM;
+       newinfo = xt_alloc_table_info(size);
+       if (!newinfo)
+               goto out_unlock;
 
-       duprintf("table->private->number = %u\n",
-                table->private->number);
-       
-       /* save number of initial entries */
-       table->private->initial_entries = table->private->number;
+       newinfo->number = number;
+       for (i = 0; i < NF_IP_NUMHOOKS; i++) {
+               newinfo->hook_entry[i] = info->hook_entry[i];
+               newinfo->underflow[i] = info->underflow[i];
+       }
+       entry1 = newinfo->entries[raw_smp_processor_id()];
+       pos = entry1;
+       size =  total_size;
+       ret = IPT_ENTRY_ITERATE(entry0, total_size,
+                       compat_copy_entry_from_user, &pos, &size,
+                       name, newinfo, entry1);
+       compat_flush_offsets();
+       xt_compat_unlock(AF_INET);
+       if (ret)
+               goto free_newinfo;
+
+       ret = -ELOOP;
+       if (!mark_source_chains(newinfo, valid_hooks, entry1))
+               goto free_newinfo;
 
-       rwlock_init(&table->lock);
-       list_prepend(&ipt_tables, table);
+       /* And one copy for every other CPU */
+       for_each_cpu(i)
+               if (newinfo->entries[i] && newinfo->entries[i] != entry1)
+                       memcpy(newinfo->entries[i], entry1, newinfo->size);
 
- unlock:
-       up(&ipt_mutex);
-       return ret;
+       *pinfo = newinfo;
+       *pentry0 = entry1;
+       xt_free_table_info(info);
+       return 0;
 
- free_unlock:
-       vfree(newinfo);
-       goto unlock;
+free_newinfo:
+       xt_free_table_info(newinfo);
+out:
+       return ret;
+out_unlock:
+       xt_compat_unlock(AF_INET);
+       goto out;
 }
 
-void ipt_unregister_table(struct ipt_table *table)
+static int
+compat_do_replace(void __user *user, unsigned int len)
 {
-       down(&ipt_mutex);
-       LIST_DELETE(&ipt_tables, table);
-       up(&ipt_mutex);
+       int ret;
+       struct compat_ipt_replace tmp;
+       struct xt_table_info *newinfo;
+       void *loc_cpu_entry;
 
-       /* Decrease module usage counts and free resources */
-       IPT_ENTRY_ITERATE(table->private->entries, table->private->size,
-                         cleanup_entry, NULL);
-       vfree(table->private);
+       if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
+               return -EFAULT;
+
+       /* Hack: Causes ipchains to give correct error msg --RR */
+       if (len != sizeof(tmp) + tmp.size)
+               return -ENOPROTOOPT;
+
+       /* overflow check */
+       if (tmp.size >= (INT_MAX - sizeof(struct xt_table_info)) / NR_CPUS -
+                       SMP_CACHE_BYTES)
+               return -ENOMEM;
+       if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
+               return -ENOMEM;
+
+       newinfo = xt_alloc_table_info(tmp.size);
+       if (!newinfo)
+               return -ENOMEM;
+
+       /* choose the copy that is our node/cpu */
+       loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
+       if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
+                          tmp.size) != 0) {
+               ret = -EFAULT;
+               goto free_newinfo;
+       }
+
+       ret = translate_compat_table(tmp.name, tmp.valid_hooks,
+                             &newinfo, &loc_cpu_entry, tmp.size,
+                             tmp.num_entries, tmp.hook_entry, tmp.underflow);
+       if (ret != 0)
+               goto free_newinfo;
+
+       duprintf("compat_do_replace: Translated table\n");
+
+       ret = __do_replace(tmp.name, tmp.valid_hooks,
+                             newinfo, tmp.num_counters,
+                             compat_ptr(tmp.counters));
+       if (ret)
+               goto free_newinfo_untrans;
+       return 0;
+
+ free_newinfo_untrans:
+       IPT_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry,NULL);
+ free_newinfo:
+       xt_free_table_info(newinfo);
+       return ret;
 }
 
-/* Returns 1 if the port is matched by the range, 0 otherwise */
-static inline int
-port_match(u_int16_t min, u_int16_t max, u_int16_t port, int invert)
+static int
+compat_do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user,
+               unsigned int len)
 {
        int ret;
 
-       ret = (port >= min && port <= max) ^ invert;
+       if (!capable(CAP_NET_ADMIN))
+               return -EPERM;
+
+       switch (cmd) {
+       case IPT_SO_SET_REPLACE:
+               ret = compat_do_replace(user, len);
+               break;
+
+       case IPT_SO_SET_ADD_COUNTERS:
+               ret = do_add_counters(user, len, 1);
+               break;
+
+       default:
+               duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
+               ret = -EINVAL;
+       }
+
        return ret;
 }
 
-static int
-tcp_find_option(u_int8_t option,
-               const struct sk_buff *skb,
-               unsigned int optlen,
-               int invert,
-               int *hotdrop)
+struct compat_ipt_get_entries
 {
-       /* tcp.doff is only 4 bits, ie. max 15 * 4 bytes */
-       u_int8_t _opt[60 - sizeof(struct tcphdr)], *op;
-       unsigned int i;
+       char name[IPT_TABLE_MAXNAMELEN];
+       compat_uint_t size;
+       struct compat_ipt_entry entrytable[0];
+};
 
-       duprintf("tcp_match: finding option\n");
+static int compat_copy_entries_to_user(unsigned int total_size,
+                    struct ipt_table *table, void __user *userptr)
+{
+       unsigned int off, num;
+       struct compat_ipt_entry e;
+       struct xt_counters *counters;
+       struct xt_table_info *private = table->private;
+       void __user *pos;
+       unsigned int size;
+       int ret = 0;
+       void *loc_cpu_entry;
+
+       counters = alloc_counters(table);
+       if (IS_ERR(counters))
+               return PTR_ERR(counters);
+
+       /* choose the copy that is on our node/cpu, ...
+        * This choice is lazy (because current thread is
+        * allowed to migrate to another cpu)
+        */
+       loc_cpu_entry = private->entries[raw_smp_processor_id()];
+       pos = userptr;
+       size = total_size;
+       ret = IPT_ENTRY_ITERATE(loc_cpu_entry, total_size,
+                       compat_copy_entry_to_user, &pos, &size);
+       if (ret)
+               goto free_counters;
 
-       if (!optlen)
-               return invert;
+       /* ... then go back and fix counters and names */
+       for (off = 0, num = 0; off < size; off += e.next_offset, num++) {
+               unsigned int i;
+               struct ipt_entry_match m;
+               struct ipt_entry_target t;
 
-       /* If we don't have the whole header, drop packet. */
-       op = skb_header_pointer(skb,
-                               skb->nh.iph->ihl*4 + sizeof(struct tcphdr),
-                               optlen, _opt);
-       if (op == NULL) {
-               *hotdrop = 1;
-               return 0;
-       }
+               ret = -EFAULT;
+               if (copy_from_user(&e, userptr + off,
+                                       sizeof(struct compat_ipt_entry)))
+                       goto free_counters;
+               if (copy_to_user(userptr + off +
+                       offsetof(struct compat_ipt_entry, counters),
+                        &counters[num], sizeof(counters[num])))
+                       goto free_counters;
 
-       for (i = 0; i < optlen; ) {
-               if (op[i] == option) return !invert;
-               if (op[i] < 2) i++;
-               else i += op[i+1]?:1;
-       }
+               for (i = sizeof(struct compat_ipt_entry);
+                               i < e.target_offset; i += m.u.match_size) {
+                       if (copy_from_user(&m, userptr + off + i,
+                                       sizeof(struct ipt_entry_match)))
+                               goto free_counters;
+                       if (copy_to_user(userptr + off + i +
+                               offsetof(struct ipt_entry_match, u.user.name),
+                               m.u.kernel.match->name,
+                               strlen(m.u.kernel.match->name) + 1))
+                               goto free_counters;
+               }
 
-       return invert;
+               if (copy_from_user(&t, userptr + off + e.target_offset,
+                                       sizeof(struct ipt_entry_target)))
+                       goto free_counters;
+               if (copy_to_user(userptr + off + e.target_offset +
+                       offsetof(struct ipt_entry_target, u.user.name),
+                       t.u.kernel.target->name,
+                       strlen(t.u.kernel.target->name) + 1))
+                       goto free_counters;
+       }
+       ret = 0;
+free_counters:
+       vfree(counters);
+       return ret;
 }
 
 static int
-tcp_match(const struct sk_buff *skb,
-         const struct net_device *in,
-         const struct net_device *out,
-         const void *matchinfo,
-         int offset,
-         int *hotdrop)
+compat_get_entries(struct compat_ipt_get_entries __user *uptr, int *len)
 {
-       struct tcphdr _tcph, *th;
-       const struct ipt_tcp *tcpinfo = matchinfo;
-
-       if (offset) {
-               /* To quote Alan:
-
-                  Don't allow a fragment of TCP 8 bytes in. Nobody normal
-                  causes this. Its a cracker trying to break in by doing a
-                  flag overwrite to pass the direction checks.
-               */
-               if (offset == 1) {
-                       duprintf("Dropping evil TCP offset=1 frag.\n");
-                       *hotdrop = 1;
-               }
-               /* Must not be a fragment. */
-               return 0;
+       int ret;
+       struct compat_ipt_get_entries get;
+       struct ipt_table *t;
+
+
+       if (*len < sizeof(get)) {
+               duprintf("compat_get_entries: %u < %u\n",
+                               *len, (unsigned int)sizeof(get));
+               return -EINVAL;
        }
 
-#define FWINVTCP(bool,invflg) ((bool) ^ !!(tcpinfo->invflags & invflg))
+       if (copy_from_user(&get, uptr, sizeof(get)) != 0)
+               return -EFAULT;
 
-       th = skb_header_pointer(skb, skb->nh.iph->ihl*4,
-                               sizeof(_tcph), &_tcph);
-       if (th == NULL) {
-               /* We've been asked to examine this packet, and we
-                  can't.  Hence, no choice but to drop. */
-               duprintf("Dropping evil TCP offset=0 tinygram.\n");
-               *hotdrop = 1;
-               return 0;
+       if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
+               duprintf("compat_get_entries: %u != %u\n", *len,
+                       (unsigned int)(sizeof(struct compat_ipt_get_entries) +
+                       get.size));
+               return -EINVAL;
        }
 
-       if (!port_match(tcpinfo->spts[0], tcpinfo->spts[1],
-                       ntohs(th->source),
-                       !!(tcpinfo->invflags & IPT_TCP_INV_SRCPT)))
-               return 0;
-       if (!port_match(tcpinfo->dpts[0], tcpinfo->dpts[1],
-                       ntohs(th->dest),
-                       !!(tcpinfo->invflags & IPT_TCP_INV_DSTPT)))
-               return 0;
-       if (!FWINVTCP((((unsigned char *)th)[13] & tcpinfo->flg_mask)
-                     == tcpinfo->flg_cmp,
-                     IPT_TCP_INV_FLAGS))
-               return 0;
-       if (tcpinfo->option) {
-               if (th->doff * 4 < sizeof(_tcph)) {
-                       *hotdrop = 1;
-                       return 0;
+       xt_compat_lock(AF_INET);
+       t = xt_find_table_lock(AF_INET, get.name);
+       if (t && !IS_ERR(t)) {
+               struct xt_table_info *private = t->private;
+               struct xt_table_info info;
+               duprintf("t->private->number = %u\n",
+                        private->number);
+               ret = compat_table_info(private, &info);
+               if (!ret && get.size == info.size) {
+                       ret = compat_copy_entries_to_user(private->size,
+                                                  t, uptr->entrytable);
+               } else if (!ret) {
+                       duprintf("compat_get_entries: I've got %u not %u!\n",
+                                private->size,
+                                get.size);
+                       ret = -EINVAL;
                }
-               if (!tcp_find_option(tcpinfo->option, skb,
-                                    th->doff*4 - sizeof(_tcph),
-                                    tcpinfo->invflags & IPT_TCP_INV_OPTION,
-                                    hotdrop))
-                       return 0;
-       }
-       return 1;
+               compat_flush_offsets();
+               module_put(t->me);
+               xt_table_unlock(t);
+       } else
+               ret = t ? PTR_ERR(t) : -ENOENT;
+
+       xt_compat_unlock(AF_INET);
+       return ret;
 }
 
-/* Called when user tries to insert an entry of this type. */
 static int
-tcp_checkentry(const char *tablename,
-              const struct ipt_ip *ip,
-              void *matchinfo,
-              unsigned int matchsize,
-              unsigned int hook_mask)
+compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
 {
-       const struct ipt_tcp *tcpinfo = matchinfo;
+       int ret;
 
-       /* Must specify proto == TCP, and no unknown invflags */
-       return ip->proto == IPPROTO_TCP
-               && !(ip->invflags & IPT_INV_PROTO)
-               && matchsize == IPT_ALIGN(sizeof(struct ipt_tcp))
-               && !(tcpinfo->invflags & ~IPT_TCP_INV_MASK);
+       switch (cmd) {
+       case IPT_SO_GET_INFO:
+               ret = get_info(user, len, 1);
+               break;
+       case IPT_SO_GET_ENTRIES:
+               ret = compat_get_entries(user, len);
+               break;
+       default:
+               duprintf("compat_do_ipt_get_ctl: unknown request %i\n", cmd);
+               ret = -EINVAL;
+       }
+       return ret;
 }
+#endif
 
 static int
-udp_match(const struct sk_buff *skb,
-         const struct net_device *in,
-         const struct net_device *out,
-         const void *matchinfo,
-         int offset,
-         int *hotdrop)
+do_ipt_set_ctl(struct sock *sk,        int cmd, void __user *user, unsigned int len)
 {
-       struct udphdr _udph, *uh;
-       const struct ipt_udp *udpinfo = matchinfo;
+       int ret;
 
-       /* Must not be a fragment. */
-       if (offset)
-               return 0;
+       if (!capable(CAP_NET_ADMIN))
+               return -EPERM;
 
-       uh = skb_header_pointer(skb, skb->nh.iph->ihl*4,
-                               sizeof(_udph), &_udph);
-       if (uh == NULL) {
-               /* We've been asked to examine this packet, and we
-                  can't.  Hence, no choice but to drop. */
-               duprintf("Dropping evil UDP tinygram.\n");
-               *hotdrop = 1;
-               return 0;
+       switch (cmd) {
+       case IPT_SO_SET_REPLACE:
+               ret = do_replace(user, len);
+               break;
+
+       case IPT_SO_SET_ADD_COUNTERS:
+               ret = do_add_counters(user, len, 0);
+               break;
+
+       default:
+               duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
+               ret = -EINVAL;
        }
 
-       return port_match(udpinfo->spts[0], udpinfo->spts[1],
-                         ntohs(uh->source),
-                         !!(udpinfo->invflags & IPT_UDP_INV_SRCPT))
-               && port_match(udpinfo->dpts[0], udpinfo->dpts[1],
-                             ntohs(uh->dest),
-                             !!(udpinfo->invflags & IPT_UDP_INV_DSTPT));
+       return ret;
 }
 
-/* Called when user tries to insert an entry of this type. */
 static int
-udp_checkentry(const char *tablename,
-              const struct ipt_ip *ip,
-              void *matchinfo,
-              unsigned int matchinfosize,
-              unsigned int hook_mask)
+do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
 {
-       const struct ipt_udp *udpinfo = matchinfo;
+       int ret;
 
-       /* Must specify proto == UDP, and no unknown invflags */
-       if (ip->proto != IPPROTO_UDP || (ip->invflags & IPT_INV_PROTO)) {
-               duprintf("ipt_udp: Protocol %u != %u\n", ip->proto,
-                        IPPROTO_UDP);
-               return 0;
+       if (!capable(CAP_NET_ADMIN))
+               return -EPERM;
+
+       switch (cmd) {
+       case IPT_SO_GET_INFO:
+               ret = get_info(user, len, 0);
+               break;
+
+       case IPT_SO_GET_ENTRIES:
+               ret = get_entries(user, len);
+               break;
+
+       case IPT_SO_GET_REVISION_MATCH:
+       case IPT_SO_GET_REVISION_TARGET: {
+               struct ipt_get_revision rev;
+               int target;
+
+               if (*len != sizeof(rev)) {
+                       ret = -EINVAL;
+                       break;
+               }
+               if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
+                       ret = -EFAULT;
+                       break;
+               }
+
+               if (cmd == IPT_SO_GET_REVISION_TARGET)
+                       target = 1;
+               else
+                       target = 0;
+
+               try_then_request_module(xt_find_revision(AF_INET, rev.name,
+                                                        rev.revision,
+                                                        target, &ret),
+                                       "ipt_%s", rev.name);
+               break;
        }
-       if (matchinfosize != IPT_ALIGN(sizeof(struct ipt_udp))) {
-               duprintf("ipt_udp: matchsize %u != %u\n",
-                        matchinfosize, IPT_ALIGN(sizeof(struct ipt_udp)));
-               return 0;
+
+       default:
+               duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
+               ret = -EINVAL;
        }
-       if (udpinfo->invflags & ~IPT_UDP_INV_MASK) {
-               duprintf("ipt_udp: unknown flags %X\n",
-                        udpinfo->invflags);
-               return 0;
+
+       return ret;
+}
+
+int ipt_register_table(struct xt_table *table, const struct ipt_replace *repl)
+{
+       int ret;
+       struct xt_table_info *newinfo;
+       static struct xt_table_info bootstrap
+               = { 0, 0, 0, { 0 }, { 0 }, { } };
+       void *loc_cpu_entry;
+
+       newinfo = xt_alloc_table_info(repl->size);
+       if (!newinfo)
+               return -ENOMEM;
+
+       /* choose the copy on our node/cpu
+        * but dont care of preemption
+        */
+       loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
+       memcpy(loc_cpu_entry, repl->entries, repl->size);
+
+       ret = translate_table(table->name, table->valid_hooks,
+                             newinfo, loc_cpu_entry, repl->size,
+                             repl->num_entries,
+                             repl->hook_entry,
+                             repl->underflow);
+       if (ret != 0) {
+               xt_free_table_info(newinfo);
+               return ret;
        }
 
-       return 1;
+       if (xt_register_table(table, &bootstrap, newinfo) != 0) {
+               xt_free_table_info(newinfo);
+               return ret;
+       }
+
+       return 0;
+}
+
+void ipt_unregister_table(struct ipt_table *table)
+{
+       struct xt_table_info *private;
+       void *loc_cpu_entry;
+
+       private = xt_unregister_table(table);
+
+       /* Decrease module usage counts and free resources */
+       loc_cpu_entry = private->entries[raw_smp_processor_id()];
+       IPT_ENTRY_ITERATE(loc_cpu_entry, private->size, cleanup_entry, NULL);
+       xt_free_table_info(private);
 }
 
 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
@@ -1725,8 +2148,10 @@ static int
 icmp_match(const struct sk_buff *skb,
           const struct net_device *in,
           const struct net_device *out,
+          const struct xt_match *match,
           const void *matchinfo,
           int offset,
+          unsigned int protoff,
           int *hotdrop)
 {
        struct icmphdr _icmph, *ic;
@@ -1736,8 +2161,7 @@ icmp_match(const struct sk_buff *skb,
        if (offset)
                return 0;
 
-       ic = skb_header_pointer(skb, skb->nh.iph->ihl*4,
-                               sizeof(_icmph), &_icmph);
+       ic = skb_header_pointer(skb, protoff, sizeof(_icmph), &_icmph);
        if (ic == NULL) {
                /* We've been asked to examine this packet, and we
                 * can't.  Hence, no choice but to drop.
@@ -1757,28 +2181,33 @@ icmp_match(const struct sk_buff *skb,
 /* Called when user tries to insert an entry of this type. */
 static int
 icmp_checkentry(const char *tablename,
-          const struct ipt_ip *ip,
+          const void *info,
+          const struct xt_match *match,
           void *matchinfo,
           unsigned int matchsize,
           unsigned int hook_mask)
 {
        const struct ipt_icmp *icmpinfo = matchinfo;
 
-       /* Must specify proto == ICMP, and no unknown invflags */
-       return ip->proto == IPPROTO_ICMP
-               && !(ip->invflags & IPT_INV_PROTO)
-               && matchsize == IPT_ALIGN(sizeof(struct ipt_icmp))
-               && !(icmpinfo->invflags & ~IPT_ICMP_INV);
+       /* Must specify no unknown invflags */
+       return !(icmpinfo->invflags & ~IPT_ICMP_INV);
 }
 
 /* The built-in targets: standard (NULL) and error. */
 static struct ipt_target ipt_standard_target = {
        .name           = IPT_STANDARD_TARGET,
+       .targetsize     = sizeof(int),
+       .family         = AF_INET,
+#ifdef CONFIG_COMPAT
+       .compat         = &compat_ipt_standard_fn,
+#endif
 };
 
 static struct ipt_target ipt_error_target = {
        .name           = IPT_ERROR_TARGET,
        .target         = ipt_error,
+       .targetsize     = IPT_FUNCTION_MAXNAMELEN,
+       .family         = AF_INET,
 };
 
 static struct nf_sockopt_ops ipt_sockopts = {
@@ -1786,128 +2215,36 @@ static struct nf_sockopt_ops ipt_sockopts = {
        .set_optmin     = IPT_BASE_CTL,
        .set_optmax     = IPT_SO_SET_MAX+1,
        .set            = do_ipt_set_ctl,
+#ifdef CONFIG_COMPAT
+       .compat_set     = compat_do_ipt_set_ctl,
+#endif
        .get_optmin     = IPT_BASE_CTL,
        .get_optmax     = IPT_SO_GET_MAX+1,
        .get            = do_ipt_get_ctl,
-};
-
-static struct ipt_match tcp_matchstruct = {
-       .name           = "tcp",
-       .match          = &tcp_match,
-       .checkentry     = &tcp_checkentry,
-};
-
-static struct ipt_match udp_matchstruct = {
-       .name           = "udp",
-       .match          = &udp_match,
-       .checkentry     = &udp_checkentry,
+#ifdef CONFIG_COMPAT
+       .compat_get     = compat_do_ipt_get_ctl,
+#endif
 };
 
 static struct ipt_match icmp_matchstruct = {
        .name           = "icmp",
-       .match          = &icmp_match,
-       .checkentry     = &icmp_checkentry,
+       .match          = icmp_match,
+       .matchsize      = sizeof(struct ipt_icmp),
+       .proto          = IPPROTO_ICMP,
+       .family         = AF_INET,
+       .checkentry     = icmp_checkentry,
 };
 
-#ifdef CONFIG_PROC_FS
-static inline int print_name(const char *i,
-                            off_t start_offset, char *buffer, int length,
-                            off_t *pos, unsigned int *count)
-{
-       if ((*count)++ >= start_offset) {
-               unsigned int namelen;
-
-               namelen = sprintf(buffer + *pos, "%s\n",
-                                 i + sizeof(struct list_head));
-               if (*pos + namelen > length) {
-                       /* Stop iterating */
-                       return 1;
-               }
-               *pos += namelen;
-       }
-       return 0;
-}
-
-static inline int print_target(const struct ipt_target *t,
-                               off_t start_offset, char *buffer, int length,
-                               off_t *pos, unsigned int *count)
-{
-       if (t == &ipt_standard_target || t == &ipt_error_target)
-               return 0;
-       return print_name((char *)t, start_offset, buffer, length, pos, count);
-}
-
-static int ipt_get_tables(char *buffer, char **start, off_t offset, int length)
-{
-       off_t pos = 0;
-       unsigned int count = 0;
-
-       if (down_interruptible(&ipt_mutex) != 0)
-               return 0;
-
-       LIST_FIND(&ipt_tables, print_name, void *,
-                 offset, buffer, length, &pos, &count);
-
-       up(&ipt_mutex);
-
-       /* `start' hack - see fs/proc/generic.c line ~105 */
-       *start=(char *)((unsigned long)count-offset);
-       return pos;
-}
-
-static int ipt_get_targets(char *buffer, char **start, off_t offset, int length)
-{
-       off_t pos = 0;
-       unsigned int count = 0;
-
-       if (down_interruptible(&ipt_mutex) != 0)
-               return 0;
-
-       LIST_FIND(&ipt_target, print_target, struct ipt_target *,
-                 offset, buffer, length, &pos, &count);
-       
-       up(&ipt_mutex);
-
-       *start = (char *)((unsigned long)count - offset);
-       return pos;
-}
-
-static int ipt_get_matches(char *buffer, char **start, off_t offset, int length)
-{
-       off_t pos = 0;
-       unsigned int count = 0;
-
-       if (down_interruptible(&ipt_mutex) != 0)
-               return 0;
-       
-       LIST_FIND(&ipt_match, print_name, void *,
-                 offset, buffer, length, &pos, &count);
-
-       up(&ipt_mutex);
-
-       *start = (char *)((unsigned long)count - offset);
-       return pos;
-}
-
-static struct { char *name; get_info_t *get_info; } ipt_proc_entry[] =
-{ { "ip_tables_names", ipt_get_tables },
-  { "ip_tables_targets", ipt_get_targets },
-  { "ip_tables_matches", ipt_get_matches },
-  { NULL, NULL} };
-#endif /*CONFIG_PROC_FS*/
-
-static int __init init(void)
+static int __init ip_tables_init(void)
 {
        int ret;
 
+       xt_proto_init(AF_INET);
+
        /* Noone else will be downing sem now, so we won't sleep */
-       down(&ipt_mutex);
-       list_append(&ipt_target, &ipt_standard_target);
-       list_append(&ipt_target, &ipt_error_target);
-       list_append(&ipt_match, &tcp_matchstruct);
-       list_append(&ipt_match, &udp_matchstruct);
-       list_append(&ipt_match, &icmp_matchstruct);
-       up(&ipt_mutex);
+       xt_register_target(&ipt_standard_target);
+       xt_register_target(&ipt_error_target);
+       xt_register_match(&icmp_matchstruct);
 
        /* Register setsockopt */
        ret = nf_register_sockopt(&ipt_sockopts);
@@ -1916,49 +2253,23 @@ static int __init init(void)
                return ret;
        }
 
-#ifdef CONFIG_PROC_FS
-       {
-       struct proc_dir_entry *proc;
-       int i;
-
-       for (i = 0; ipt_proc_entry[i].name; i++) {
-               proc = proc_net_create(ipt_proc_entry[i].name, 0,
-                                      ipt_proc_entry[i].get_info);
-               if (!proc) {
-                       while (--i >= 0)
-                               proc_net_remove(ipt_proc_entry[i].name);
-                       nf_unregister_sockopt(&ipt_sockopts);
-                       return -ENOMEM;
-               }
-               proc->owner = THIS_MODULE;
-       }
-       }
-#endif
-
-       printk("ip_tables: (C) 2000-2002 Netfilter core team\n");
+       printk("ip_tables: (C) 2000-2006 Netfilter Core Team\n");
        return 0;
 }
 
-static void __exit fini(void)
+static void __exit ip_tables_fini(void)
 {
        nf_unregister_sockopt(&ipt_sockopts);
-#ifdef CONFIG_PROC_FS
-       {
-       int i;
-       for (i = 0; ipt_proc_entry[i].name; i++)
-               proc_net_remove(ipt_proc_entry[i].name);
-       }
-#endif
+
+       xt_unregister_match(&icmp_matchstruct);
+       xt_unregister_target(&ipt_error_target);
+       xt_unregister_target(&ipt_standard_target);
+
+       xt_proto_fini(AF_INET);
 }
 
 EXPORT_SYMBOL(ipt_register_table);
 EXPORT_SYMBOL(ipt_unregister_table);
-EXPORT_SYMBOL(ipt_register_match);
-EXPORT_SYMBOL(ipt_unregister_match);
 EXPORT_SYMBOL(ipt_do_table);
-EXPORT_SYMBOL(ipt_register_target);
-EXPORT_SYMBOL(ipt_unregister_target);
-EXPORT_SYMBOL(ipt_find_target);
-
-module_init(init);
-module_exit(fini);
+module_init(ip_tables_init);
+module_exit(ip_tables_fini);