fedora core 6 1.2949 + vserver 2.2.0
[linux-2.6.git] / net / ipv4 / netfilter / ipt_MASQUERADE.c
index fb49990..d669685 100644 (file)
@@ -2,23 +2,29 @@
    (depending on route). */
 
 /* (C) 1999-2001 Paul `Rusty' Russell
- * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
+ * (C) 2002-2006 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
  * published by the Free Software Foundation.
  */
 
-#include <linux/config.h>
 #include <linux/types.h>
+#include <linux/inetdevice.h>
 #include <linux/ip.h>
 #include <linux/timer.h>
 #include <linux/module.h>
 #include <linux/netfilter.h>
 #include <net/protocol.h>
+#include <net/ip.h>
 #include <net/checksum.h>
+#include <net/route.h>
 #include <linux/netfilter_ipv4.h>
+#ifdef CONFIG_NF_NAT_NEEDED
+#include <net/netfilter/nf_nat_rule.h>
+#else
 #include <linux/netfilter_ipv4/ip_nat_rule.h>
+#endif
 #include <linux/netfilter_ipv4/ip_tables.h>
 
 MODULE_LICENSE("GPL");
@@ -32,31 +38,18 @@ MODULE_DESCRIPTION("iptables MASQUERADE target module");
 #endif
 
 /* Lock protects masq region inside conntrack */
-static DECLARE_RWLOCK(masq_lock);
+static DEFINE_RWLOCK(masq_lock);
 
 /* FIXME: Multiple targets. --RR */
 static int
 masquerade_check(const char *tablename,
-                const struct ipt_entry *e,
+                const void *e,
+                const struct xt_target *target,
                 void *targinfo,
-                unsigned int targinfosize,
                 unsigned int hook_mask)
 {
-       const struct ip_nat_multi_range *mr = targinfo;
+       const struct ip_nat_multi_range_compat *mr = targinfo;
 
-       if (strcmp(tablename, "nat") != 0) {
-               DEBUGP("masquerade_check: bad table `%s'.\n", tablename);
-               return 0;
-       }
-       if (targinfosize != IPT_ALIGN(sizeof(*mr))) {
-               DEBUGP("masquerade_check: size %u != %u.\n",
-                      targinfosize, sizeof(*mr));
-               return 0;
-       }
-       if (hook_mask & ~(1 << NF_IP_POST_ROUTING)) {
-               DEBUGP("masquerade_check: bad hooks %x.\n", hook_mask);
-               return 0;
-       }
        if (mr->range[0].flags & IP_NAT_RANGE_MAP_IPS) {
                DEBUGP("masquerade_check: bad MAP_IPS.\n");
                return 0;
@@ -73,102 +66,126 @@ masquerade_target(struct sk_buff **pskb,
                  const struct net_device *in,
                  const struct net_device *out,
                  unsigned int hooknum,
-                 const void *targinfo,
-                 void *userinfo)
+                 const struct xt_target *target,
+                 const void *targinfo)
 {
+#ifdef CONFIG_NF_NAT_NEEDED
+       struct nf_conn_nat *nat;
+#endif
        struct ip_conntrack *ct;
        enum ip_conntrack_info ctinfo;
-       const struct ip_nat_multi_range *mr;
-       struct ip_nat_multi_range newrange;
-       u_int32_t newsrc;
+       struct ip_nat_range newrange;
+       const struct ip_nat_multi_range_compat *mr;
        struct rtable *rt;
+       __be32 newsrc;
 
        IP_NF_ASSERT(hooknum == NF_IP_POST_ROUTING);
 
-       /* FIXME: For the moment, don't do local packets, breaks
-          testsuite for 2.3.49 --RR */
-       if ((*pskb)->sk)
-               return NF_ACCEPT;
-
        ct = ip_conntrack_get(*pskb, &ctinfo);
-       IP_NF_ASSERT(ct && (ctinfo == IP_CT_NEW
-                                 || ctinfo == IP_CT_RELATED));
+#ifdef CONFIG_NF_NAT_NEEDED
+       nat = nfct_nat(ct);
+#endif
+       IP_NF_ASSERT(ct && (ctinfo == IP_CT_NEW || ctinfo == IP_CT_RELATED
+                           || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY));
+
+       /* Source address is 0.0.0.0 - locally generated packet that is
+        * probably not supposed to be masqueraded.
+        */
+#ifdef CONFIG_NF_NAT_NEEDED
+       if (ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.u3.ip == 0)
+#else
+       if (ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.src.ip == 0)
+#endif
+               return NF_ACCEPT;
 
        mr = targinfo;
-
-       {
-               struct flowi fl = { .nl_u = { .ip4_u =
-                                             { .daddr = (*pskb)->nh.iph->daddr,
-                                               .tos = (RT_TOS((*pskb)->nh.iph->tos) |
-                                                       RTO_CONN),
-#ifdef CONFIG_IP_ROUTE_FWMARK
-                                               .fwmark = (*pskb)->nfmark
-#endif
-                                             } } };
-               if (ip_route_output_key(&rt, &fl) != 0) {
-                       /* Funky routing can do this. */
-                       if (net_ratelimit())
-                               printk("MASQUERADE:"
-                                      " No route: Rusty's brain broke!\n");
-                       return NF_DROP;
-               }
-               if (rt->u.dst.dev != out) {
-                       if (net_ratelimit())
-                               printk("MASQUERADE:"
-                                      " Route sent us somewhere else.\n");
-                       ip_rt_put(rt);
-                       return NF_DROP;
-               }
+       rt = (struct rtable *)(*pskb)->dst;
+       newsrc = inet_select_addr(out, rt->rt_gateway, RT_SCOPE_UNIVERSE);
+       if (!newsrc) {
+               printk("MASQUERADE: %s ate my IP address\n", out->name);
+               return NF_DROP;
        }
 
-       newsrc = rt->rt_src;
-       DEBUGP("newsrc = %u.%u.%u.%u\n", NIPQUAD(newsrc));
-       ip_rt_put(rt);
-
-       WRITE_LOCK(&masq_lock);
+       write_lock_bh(&masq_lock);
+#ifdef CONFIG_NF_NAT_NEEDED
+       nat->masq_index = out->ifindex;
+#else
        ct->nat.masq_index = out->ifindex;
-       WRITE_UNLOCK(&masq_lock);
+#endif
+       write_unlock_bh(&masq_lock);
 
        /* Transfer from original range. */
-       newrange = ((struct ip_nat_multi_range)
-               { 1, { { mr->range[0].flags | IP_NAT_RANGE_MAP_IPS,
-                        newsrc, newsrc,
-                        mr->range[0].min, mr->range[0].max } } });
+       newrange = ((struct ip_nat_range)
+               { mr->range[0].flags | IP_NAT_RANGE_MAP_IPS,
+                 newsrc, newsrc,
+                 mr->range[0].min, mr->range[0].max });
 
        /* Hand modified range to generic setup. */
        return ip_nat_setup_info(ct, &newrange, hooknum);
 }
 
 static inline int
-device_cmp(const struct ip_conntrack *i, void *_ina)
+device_cmp(struct ip_conntrack *i, void *ifindex)
 {
-       int ret = 0;
-       struct in_ifaddr *ina = _ina;
+       int ret;
+#ifdef CONFIG_NF_NAT_NEEDED
+       struct nf_conn_nat *nat = nfct_nat(i);
 
-       READ_LOCK(&masq_lock);
-       /* If it's masquerading out this interface with a different address,
-          or we don't know the new address of this interface. */
-       if (i->nat.masq_index == ina->ifa_dev->dev->ifindex
-           && i->tuplehash[IP_CT_DIR_REPLY].tuple.dst.ip != ina->ifa_address)
-               ret = 1;
-       READ_UNLOCK(&masq_lock);
+       if (!nat)
+               return 0;
+#endif
+
+       read_lock_bh(&masq_lock);
+#ifdef CONFIG_NF_NAT_NEEDED
+       ret = (nat->masq_index == (int)(long)ifindex);
+#else
+       ret = (i->nat.masq_index == (int)(long)ifindex);
+#endif
+       read_unlock_bh(&masq_lock);
 
        return ret;
 }
 
+static int masq_device_event(struct notifier_block *this,
+                            unsigned long event,
+                            void *ptr)
+{
+       struct net_device *dev = ptr;
+
+       if (event == NETDEV_DOWN) {
+               /* Device was downed.  Search entire table for
+                  conntracks which were associated with that device,
+                  and forget them. */
+               IP_NF_ASSERT(dev->ifindex != 0);
+
+               ip_ct_iterate_cleanup(device_cmp, (void *)(long)dev->ifindex);
+       }
+
+       return NOTIFY_DONE;
+}
+
 static int masq_inet_event(struct notifier_block *this,
                           unsigned long event,
                           void *ptr)
 {
-       /* For some configurations, interfaces often come back with
-        * the same address.  If not, clean up old conntrack
-        * entries. */
-       if (event == NETDEV_UP)
-               ip_ct_selective_cleanup(device_cmp, ptr);
+       struct net_device *dev = ((struct in_ifaddr *)ptr)->ifa_dev->dev;
+
+       if (event == NETDEV_DOWN) {
+               /* IP address was deleted.  Search entire table for
+                  conntracks which were associated with that device,
+                  and forget them. */
+               IP_NF_ASSERT(dev->ifindex != 0);
+
+               ip_ct_iterate_cleanup(device_cmp, (void *)(long)dev->ifindex);
+       }
 
        return NOTIFY_DONE;
 }
 
+static struct notifier_block masq_dev_notifier = {
+       .notifier_call  = masq_device_event,
+};
+
 static struct notifier_block masq_inet_notifier = {
        .notifier_call  = masq_inet_event,
 };
@@ -176,28 +193,35 @@ static struct notifier_block masq_inet_notifier = {
 static struct ipt_target masquerade = {
        .name           = "MASQUERADE",
        .target         = masquerade_target,
+       .targetsize     = sizeof(struct ip_nat_multi_range_compat),
+       .table          = "nat",
+       .hooks          = 1 << NF_IP_POST_ROUTING,
        .checkentry     = masquerade_check,
        .me             = THIS_MODULE,
 };
 
-static int __init init(void)
+static int __init ipt_masquerade_init(void)
 {
        int ret;
 
        ret = ipt_register_target(&masquerade);
 
-       if (ret == 0)
+       if (ret == 0) {
+               /* Register for device down reports */
+               register_netdevice_notifier(&masq_dev_notifier);
                /* Register IP address change reports */
                register_inetaddr_notifier(&masq_inet_notifier);
+       }
 
        return ret;
 }
 
-static void __exit fini(void)
+static void __exit ipt_masquerade_fini(void)
 {
        ipt_unregister_target(&masquerade);
+       unregister_netdevice_notifier(&masq_dev_notifier);
        unregister_inetaddr_notifier(&masq_inet_notifier);      
 }
 
-module_init(init);
-module_exit(fini);
+module_init(ipt_masquerade_init);
+module_exit(ipt_masquerade_fini);