netfilter: netns nf_conntrack: per-netns net.netfilter.nf_conntrack_log_invalid sysctl
[safe/jmp/linux-2.6] / net / ipv4 / netfilter / nf_conntrack_l3proto_ipv4.c
index 9bdbb77..af69acc 100644 (file)
@@ -4,17 +4,8 @@
  * 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.
- *
- * 16 Dec 2003: Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp>
- *     - move L3 protocol dependent part to this file.
- * 23 Mar 2004: Yasuyuki Kozakai @USAGI <yasuyuki.kozakai@toshiba.co.jp>
- *     - add get_features() to support various size of conntrack
- *       structures.
- *
- * Derived from net/ipv4/netfilter/ip_conntrack_standalone.c
  */
 
-#include <linux/config.h>
 #include <linux/types.h>
 #include <linux/ip.h>
 #include <linux/netfilter.h>
 #include <linux/netfilter_ipv4.h>
 #include <net/netfilter/nf_conntrack.h>
 #include <net/netfilter/nf_conntrack_helper.h>
-#include <net/netfilter/nf_conntrack_protocol.h>
+#include <net/netfilter/nf_conntrack_l4proto.h>
 #include <net/netfilter/nf_conntrack_l3proto.h>
 #include <net/netfilter/nf_conntrack_core.h>
 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
+#include <net/netfilter/nf_nat_helper.h>
 
-#if 0
-#define DEBUGP printk
-#else
-#define DEBUGP(format, args...)
-#endif
+int (*nf_nat_seq_adjust_hook)(struct sk_buff *skb,
+                             struct nf_conn *ct,
+                             enum ip_conntrack_info ctinfo);
+EXPORT_SYMBOL_GPL(nf_nat_seq_adjust_hook);
 
-DECLARE_PER_CPU(struct nf_conntrack_stat, nf_conntrack_stat);
-
-static int ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
-                            struct nf_conntrack_tuple *tuple)
+static bool ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
+                             struct nf_conntrack_tuple *tuple)
 {
-       u_int32_t _addrs[2], *ap;
+       const __be32 *ap;
+       __be32 _addrs[2];
        ap = skb_header_pointer(skb, nhoff + offsetof(struct iphdr, saddr),
                                sizeof(u_int32_t) * 2, _addrs);
        if (ap == NULL)
-               return 0;
+               return false;
 
        tuple->src.u3.ip = ap[0];
        tuple->dst.u3.ip = ap[1];
 
-       return 1;
+       return true;
 }
 
-static int ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
-                          const struct nf_conntrack_tuple *orig)
+static bool ipv4_invert_tuple(struct nf_conntrack_tuple *tuple,
+                             const struct nf_conntrack_tuple *orig)
 {
        tuple->src.u3.ip = orig->dst.u3.ip;
        tuple->dst.u3.ip = orig->src.u3.ip;
 
-       return 1;
+       return true;
 }
 
 static int ipv4_print_tuple(struct seq_file *s,
                            const struct nf_conntrack_tuple *tuple)
 {
        return seq_printf(s, "src=%u.%u.%u.%u dst=%u.%u.%u.%u ",
-                         NIPQUAD(tuple->src.u3.ip),
+                         NIPQUAD(tuple->src.u3.ip),
                          NIPQUAD(tuple->dst.u3.ip));
 }
 
-static int ipv4_print_conntrack(struct seq_file *s,
-                               const struct nf_conn *conntrack)
-{
-       return 0;
-}
-
 /* Returns new sk_buff, or NULL */
-static struct sk_buff *
-nf_ct_ipv4_gather_frags(struct sk_buff *skb, u_int32_t user)
+static int nf_ct_ipv4_gather_frags(struct sk_buff *skb, u_int32_t user)
 {
+       int err;
+
        skb_orphan(skb);
 
-        local_bh_disable();
-        skb = ip_defrag(skb, user);
-        local_bh_enable();
+       local_bh_disable();
+       err = ip_defrag(skb, user);
+       local_bh_enable();
 
-        if (skb)
-               ip_send_check(skb->nh.iph);
+       if (!err)
+               ip_send_check(ip_hdr(skb));
 
-        return skb;
+       return err;
 }
 
-static int
-ipv4_prepare(struct sk_buff **pskb, unsigned int hooknum, unsigned int *dataoff,
-            u_int8_t *protonum)
+static int ipv4_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
+                           unsigned int *dataoff, u_int8_t *protonum)
 {
-       /* Never happen */
-       if ((*pskb)->nh.iph->frag_off & htons(IP_OFFSET)) {
-               if (net_ratelimit()) {
-                       printk(KERN_ERR "ipv4_prepare: Frag of proto %u (hook=%u)\n",
-                       (*pskb)->nh.iph->protocol, hooknum);
-               }
-               return -NF_DROP;
-       }
+       const struct iphdr *iph;
+       struct iphdr _iph;
 
-       *dataoff = (*pskb)->nh.raw - (*pskb)->data + (*pskb)->nh.iph->ihl*4;
-       *protonum = (*pskb)->nh.iph->protocol;
+       iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
+       if (iph == NULL)
+               return -NF_DROP;
 
-       return NF_ACCEPT;
-}
+       /* Conntrack defragments packets, we might still see fragments
+        * inside ICMP packets though. */
+       if (iph->frag_off & htons(IP_OFFSET))
+               return -NF_DROP;
 
-int nat_module_is_loaded = 0;
-static u_int32_t ipv4_get_features(const struct nf_conntrack_tuple *tuple)
-{
-       if (nat_module_is_loaded)
-               return NF_CT_F_NAT;
+       *dataoff = nhoff + (iph->ihl << 2);
+       *protonum = iph->protocol;
 
-       return NF_CT_F_BASIC;
+       return NF_ACCEPT;
 }
 
 static unsigned int ipv4_confirm(unsigned int hooknum,
-                                struct sk_buff **pskb,
+                                struct sk_buff *skb,
                                 const struct net_device *in,
                                 const struct net_device *out,
                                 int (*okfn)(struct sk_buff *))
 {
-       /* We've seen it coming out the other side: confirm it */
-       return nf_conntrack_confirm(pskb);
-}
-
-static unsigned int ipv4_conntrack_help(unsigned int hooknum,
-                                     struct sk_buff **pskb,
-                                     const struct net_device *in,
-                                     const struct net_device *out,
-                                     int (*okfn)(struct sk_buff *))
-{
        struct nf_conn *ct;
        enum ip_conntrack_info ctinfo;
+       const struct nf_conn_help *help;
+       const struct nf_conntrack_helper *helper;
+       unsigned int ret;
 
        /* This is where we call the helper: as the packet goes out. */
-       ct = nf_ct_get(*pskb, &ctinfo);
-       if (ct && ct->helper) {
-               unsigned int ret;
-               ret = ct->helper->help(pskb,
-                                      (*pskb)->nh.raw - (*pskb)->data
-                                                      + (*pskb)->nh.iph->ihl*4,
-                                      ct, ctinfo);
-               if (ret != NF_ACCEPT)
-                       return ret;
+       ct = nf_ct_get(skb, &ctinfo);
+       if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
+               goto out;
+
+       help = nfct_help(ct);
+       if (!help)
+               goto out;
+
+       /* rcu_read_lock()ed by nf_hook_slow */
+       helper = rcu_dereference(help->helper);
+       if (!helper)
+               goto out;
+
+       ret = helper->help(skb, skb_network_offset(skb) + ip_hdrlen(skb),
+                          ct, ctinfo);
+       if (ret != NF_ACCEPT)
+               return ret;
+
+       if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status)) {
+               typeof(nf_nat_seq_adjust_hook) seq_adjust;
+
+               seq_adjust = rcu_dereference(nf_nat_seq_adjust_hook);
+               if (!seq_adjust || !seq_adjust(skb, ct, ctinfo))
+                       return NF_DROP;
        }
-       return NF_ACCEPT;
+out:
+       /* We've seen it coming out the other side: confirm it */
+       return nf_conntrack_confirm(skb);
 }
 
 static unsigned int ipv4_conntrack_defrag(unsigned int hooknum,
-                                         struct sk_buff **pskb,
+                                         struct sk_buff *skb,
                                          const struct net_device *in,
                                          const struct net_device *out,
                                          int (*okfn)(struct sk_buff *))
 {
-#if !defined(CONFIG_IP_NF_NAT) && !defined(CONFIG_IP_NF_NAT_MODULE)
        /* Previously seen (loopback)?  Ignore.  Do this before
           fragment check. */
-       if ((*pskb)->nfct)
+       if (skb->nfct)
                return NF_ACCEPT;
-#endif
 
        /* Gather fragments. */
-       if ((*pskb)->nh.iph->frag_off & htons(IP_MF|IP_OFFSET)) {
-               *pskb = nf_ct_ipv4_gather_frags(*pskb,
-                                               hooknum == NF_IP_PRE_ROUTING ?
-                                               IP_DEFRAG_CONNTRACK_IN :
-                                               IP_DEFRAG_CONNTRACK_OUT);
-               if (!*pskb)
+       if (ip_hdr(skb)->frag_off & htons(IP_MF | IP_OFFSET)) {
+               if (nf_ct_ipv4_gather_frags(skb,
+                                           hooknum == NF_INET_PRE_ROUTING ?
+                                           IP_DEFRAG_CONNTRACK_IN :
+                                           IP_DEFRAG_CONNTRACK_OUT))
                        return NF_STOLEN;
        }
        return NF_ACCEPT;
 }
 
 static unsigned int ipv4_conntrack_in(unsigned int hooknum,
-                                     struct sk_buff **pskb,
+                                     struct sk_buff *skb,
                                      const struct net_device *in,
                                      const struct net_device *out,
                                      int (*okfn)(struct sk_buff *))
 {
-       return nf_conntrack_in(PF_INET, hooknum, pskb);
+       return nf_conntrack_in(dev_net(in), PF_INET, hooknum, skb);
 }
 
 static unsigned int ipv4_conntrack_local(unsigned int hooknum,
-                                        struct sk_buff **pskb,
-                                        const struct net_device *in,
-                                        const struct net_device *out,
-                                        int (*okfn)(struct sk_buff *))
+                                        struct sk_buff *skb,
+                                        const struct net_device *in,
+                                        const struct net_device *out,
+                                        int (*okfn)(struct sk_buff *))
 {
        /* root is playing with raw sockets. */
-       if ((*pskb)->len < sizeof(struct iphdr)
-           || (*pskb)->nh.iph->ihl * 4 < sizeof(struct iphdr)) {
+       if (skb->len < sizeof(struct iphdr) ||
+           ip_hdrlen(skb) < sizeof(struct iphdr)) {
                if (net_ratelimit())
                        printk("ipt_hook: happy cracking.\n");
                return NF_ACCEPT;
        }
-       return nf_conntrack_in(PF_INET, hooknum, pskb);
+       return nf_conntrack_in(dev_net(out), PF_INET, hooknum, skb);
 }
 
 /* Connection tracking may drop packets, but never alters them, so
    make it the first hook. */
-static struct nf_hook_ops ipv4_conntrack_defrag_ops = {
-       .hook           = ipv4_conntrack_defrag,
-       .owner          = THIS_MODULE,
-       .pf             = PF_INET,
-       .hooknum        = NF_IP_PRE_ROUTING,
-       .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
-};
-
-static struct nf_hook_ops ipv4_conntrack_in_ops = {
-       .hook           = ipv4_conntrack_in,
-       .owner          = THIS_MODULE,
-       .pf             = PF_INET,
-       .hooknum        = NF_IP_PRE_ROUTING,
-       .priority       = NF_IP_PRI_CONNTRACK,
-};
-
-static struct nf_hook_ops ipv4_conntrack_defrag_local_out_ops = {
-       .hook           = ipv4_conntrack_defrag,
-       .owner          = THIS_MODULE,
-       .pf             = PF_INET,
-       .hooknum        = NF_IP_LOCAL_OUT,
-       .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
-};
-
-static struct nf_hook_ops ipv4_conntrack_local_out_ops = {
-       .hook           = ipv4_conntrack_local,
-       .owner          = THIS_MODULE,
-       .pf             = PF_INET,
-       .hooknum        = NF_IP_LOCAL_OUT,
-       .priority       = NF_IP_PRI_CONNTRACK,
-};
-
-/* helpers */
-static struct nf_hook_ops ipv4_conntrack_helper_out_ops = {
-       .hook           = ipv4_conntrack_help,
-       .owner          = THIS_MODULE,
-       .pf             = PF_INET,
-       .hooknum        = NF_IP_POST_ROUTING,
-       .priority       = NF_IP_PRI_CONNTRACK_HELPER,
-};
-
-static struct nf_hook_ops ipv4_conntrack_helper_in_ops = {
-       .hook           = ipv4_conntrack_help,
-       .owner          = THIS_MODULE,
-       .pf             = PF_INET,
-       .hooknum        = NF_IP_LOCAL_IN,
-       .priority       = NF_IP_PRI_CONNTRACK_HELPER,
-};
-
-
-/* Refragmenter; last chance. */
-static struct nf_hook_ops ipv4_conntrack_out_ops = {
-       .hook           = ipv4_confirm,
-       .owner          = THIS_MODULE,
-       .pf             = PF_INET,
-       .hooknum        = NF_IP_POST_ROUTING,
-       .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
-};
-
-static struct nf_hook_ops ipv4_conntrack_local_in_ops = {
-       .hook           = ipv4_confirm,
-       .owner          = THIS_MODULE,
-       .pf             = PF_INET,
-       .hooknum        = NF_IP_LOCAL_IN,
-       .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
+static struct nf_hook_ops ipv4_conntrack_ops[] __read_mostly = {
+       {
+               .hook           = ipv4_conntrack_defrag,
+               .owner          = THIS_MODULE,
+               .pf             = PF_INET,
+               .hooknum        = NF_INET_PRE_ROUTING,
+               .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
+       },
+       {
+               .hook           = ipv4_conntrack_in,
+               .owner          = THIS_MODULE,
+               .pf             = PF_INET,
+               .hooknum        = NF_INET_PRE_ROUTING,
+               .priority       = NF_IP_PRI_CONNTRACK,
+       },
+       {
+               .hook           = ipv4_conntrack_defrag,
+               .owner          = THIS_MODULE,
+               .pf             = PF_INET,
+               .hooknum        = NF_INET_LOCAL_OUT,
+               .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
+       },
+       {
+               .hook           = ipv4_conntrack_local,
+               .owner          = THIS_MODULE,
+               .pf             = PF_INET,
+               .hooknum        = NF_INET_LOCAL_OUT,
+               .priority       = NF_IP_PRI_CONNTRACK,
+       },
+       {
+               .hook           = ipv4_confirm,
+               .owner          = THIS_MODULE,
+               .pf             = PF_INET,
+               .hooknum        = NF_INET_POST_ROUTING,
+               .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
+       },
+       {
+               .hook           = ipv4_confirm,
+               .owner          = THIS_MODULE,
+               .pf             = PF_INET,
+               .hooknum        = NF_INET_LOCAL_IN,
+               .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
+       },
 };
 
-#ifdef CONFIG_SYSCTL
-/* From nf_conntrack_proto_icmp.c */
-extern unsigned long nf_ct_icmp_timeout;
-static struct ctl_table_header *nf_ct_ipv4_sysctl_header;
+#if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
+static int log_invalid_proto_min = 0;
+static int log_invalid_proto_max = 255;
 
-static ctl_table nf_ct_sysctl_table[] = {
+static ctl_table ip_ct_sysctl_table[] = {
        {
-               .ctl_name       = NET_NF_CONNTRACK_ICMP_TIMEOUT,
-               .procname       = "nf_conntrack_icmp_timeout",
-               .data           = &nf_ct_icmp_timeout,
-               .maxlen         = sizeof(unsigned int),
+               .ctl_name       = NET_IPV4_NF_CONNTRACK_MAX,
+               .procname       = "ip_conntrack_max",
+               .data           = &nf_conntrack_max,
+               .maxlen         = sizeof(int),
                .mode           = 0644,
-               .proc_handler   = &proc_dointvec_jiffies,
+               .proc_handler   = &proc_dointvec,
        },
-        { .ctl_name = 0 }
-};
-
-static ctl_table nf_ct_netfilter_table[] = {
        {
-               .ctl_name       = NET_NETFILTER,
-               .procname       = "netfilter",
-               .mode           = 0555,
-               .child          = nf_ct_sysctl_table,
+               .ctl_name       = NET_IPV4_NF_CONNTRACK_COUNT,
+               .procname       = "ip_conntrack_count",
+               .data           = &init_net.ct.count,
+               .maxlen         = sizeof(int),
+               .mode           = 0444,
+               .proc_handler   = &proc_dointvec,
+       },
+       {
+               .ctl_name       = NET_IPV4_NF_CONNTRACK_BUCKETS,
+               .procname       = "ip_conntrack_buckets",
+               .data           = &nf_conntrack_htable_size,
+               .maxlen         = sizeof(unsigned int),
+               .mode           = 0444,
+               .proc_handler   = &proc_dointvec,
        },
-       { .ctl_name = 0 }
-};
-
-static ctl_table nf_ct_net_table[] = {
        {
-               .ctl_name       = CTL_NET,
-               .procname       = "net",
-               .mode           = 0555,
-               .child          = nf_ct_netfilter_table,
+               .ctl_name       = NET_IPV4_NF_CONNTRACK_CHECKSUM,
+               .procname       = "ip_conntrack_checksum",
+               .data           = &init_net.ct.sysctl_checksum,
+               .maxlen         = sizeof(int),
+               .mode           = 0644,
+               .proc_handler   = &proc_dointvec,
+       },
+       {
+               .ctl_name       = NET_IPV4_NF_CONNTRACK_LOG_INVALID,
+               .procname       = "ip_conntrack_log_invalid",
+               .data           = &init_net.ct.sysctl_log_invalid,
+               .maxlen         = sizeof(unsigned int),
+               .mode           = 0644,
+               .proc_handler   = &proc_dointvec_minmax,
+               .strategy       = &sysctl_intvec,
+               .extra1         = &log_invalid_proto_min,
+               .extra2         = &log_invalid_proto_max,
        },
-       { .ctl_name = 0 }
+       {
+               .ctl_name       = 0
+       }
 };
-#endif
+#endif /* CONFIG_SYSCTL && CONFIG_NF_CONNTRACK_PROC_COMPAT */
 
 /* Fast function for those who don't want to parse /proc (and I don't
    blame them). */
@@ -320,11 +299,11 @@ static ctl_table nf_ct_net_table[] = {
 static int
 getorigdst(struct sock *sk, int optval, void __user *user, int *len)
 {
-       struct inet_sock *inet = inet_sk(sk);
-       struct nf_conntrack_tuple_hash *h;
+       const struct inet_sock *inet = inet_sk(sk);
+       const struct nf_conntrack_tuple_hash *h;
        struct nf_conntrack_tuple tuple;
-       
-       NF_CT_TUPLE_U_BLANK(&tuple);
+
+       memset(&tuple, 0, sizeof(tuple));
        tuple.src.u3.ip = inet->rcv_saddr;
        tuple.src.u.tcp.port = inet->sport;
        tuple.dst.u3.ip = inet->daddr;
@@ -334,17 +313,17 @@ getorigdst(struct sock *sk, int optval, void __user *user, int *len)
 
        /* We only do TCP at the moment: is there a better way? */
        if (strcmp(sk->sk_prot->name, "TCP")) {
-               DEBUGP("SO_ORIGINAL_DST: Not a TCP socket\n");
+               pr_debug("SO_ORIGINAL_DST: Not a TCP socket\n");
                return -ENOPROTOOPT;
        }
 
        if ((unsigned int) *len < sizeof(struct sockaddr_in)) {
-               DEBUGP("SO_ORIGINAL_DST: len %u not %u\n",
-                      *len, sizeof(struct sockaddr_in));
+               pr_debug("SO_ORIGINAL_DST: len %d not %Zu\n",
+                        *len, sizeof(struct sockaddr_in));
                return -EINVAL;
        }
 
-       h = nf_conntrack_find_get(&tuple, NULL);
+       h = nf_conntrack_find_get(sock_net(sk), &tuple);
        if (h) {
                struct sockaddr_in sin;
                struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
@@ -354,58 +333,51 @@ getorigdst(struct sock *sk, int optval, void __user *user, int *len)
                        .tuple.dst.u.tcp.port;
                sin.sin_addr.s_addr = ct->tuplehash[IP_CT_DIR_ORIGINAL]
                        .tuple.dst.u3.ip;
+               memset(sin.sin_zero, 0, sizeof(sin.sin_zero));
 
-               DEBUGP("SO_ORIGINAL_DST: %u.%u.%u.%u %u\n",
-                      NIPQUAD(sin.sin_addr.s_addr), ntohs(sin.sin_port));
+               pr_debug("SO_ORIGINAL_DST: %u.%u.%u.%u %u\n",
+                        NIPQUAD(sin.sin_addr.s_addr), ntohs(sin.sin_port));
                nf_ct_put(ct);
                if (copy_to_user(user, &sin, sizeof(sin)) != 0)
                        return -EFAULT;
                else
                        return 0;
        }
-       DEBUGP("SO_ORIGINAL_DST: Can't find %u.%u.%u.%u/%u-%u.%u.%u.%u/%u.\n",
-              NIPQUAD(tuple.src.u3.ip), ntohs(tuple.src.u.tcp.port),
-              NIPQUAD(tuple.dst.u3.ip), ntohs(tuple.dst.u.tcp.port));
+       pr_debug("SO_ORIGINAL_DST: Can't find %u.%u.%u.%u/%u-%u.%u.%u.%u/%u.\n",
+                NIPQUAD(tuple.src.u3.ip), ntohs(tuple.src.u.tcp.port),
+                NIPQUAD(tuple.dst.u3.ip), ntohs(tuple.dst.u.tcp.port));
        return -ENOENT;
 }
 
-#if defined(CONFIG_NF_CT_NETLINK) || \
-    defined(CONFIG_NF_CT_NETLINK_MODULE)
+#if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
 
 #include <linux/netfilter/nfnetlink.h>
 #include <linux/netfilter/nfnetlink_conntrack.h>
 
-static int ipv4_tuple_to_nfattr(struct sk_buff *skb,
+static int ipv4_tuple_to_nlattr(struct sk_buff *skb,
                                const struct nf_conntrack_tuple *tuple)
 {
-       NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t),
-               &tuple->src.u3.ip);
-       NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t),
-               &tuple->dst.u3.ip);
+       NLA_PUT_BE32(skb, CTA_IP_V4_SRC, tuple->src.u3.ip);
+       NLA_PUT_BE32(skb, CTA_IP_V4_DST, tuple->dst.u3.ip);
        return 0;
 
-nfattr_failure:
+nla_put_failure:
        return -1;
 }
 
-static const size_t cta_min_ip[CTA_IP_MAX] = {
-       [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
-       [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
+static const struct nla_policy ipv4_nla_policy[CTA_IP_MAX+1] = {
+       [CTA_IP_V4_SRC] = { .type = NLA_U32 },
+       [CTA_IP_V4_DST] = { .type = NLA_U32 },
 };
 
-static int ipv4_nfattr_to_tuple(struct nfattr *tb[],
+static int ipv4_nlattr_to_tuple(struct nlattr *tb[],
                                struct nf_conntrack_tuple *t)
 {
-       if (!tb[CTA_IP_V4_SRC-1] || !tb[CTA_IP_V4_DST-1])
-               return -EINVAL;
-
-       if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
+       if (!tb[CTA_IP_V4_SRC] || !tb[CTA_IP_V4_DST])
                return -EINVAL;
 
-       t->src.u3.ip =
-               *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
-       t->dst.u3.ip =
-               *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
+       t->src.u3.ip = nla_get_be32(tb[CTA_IP_V4_SRC]);
+       t->dst.u3.ip = nla_get_be32(tb[CTA_IP_V4_DST]);
 
        return 0;
 }
@@ -416,53 +388,60 @@ static struct nf_sockopt_ops so_getorigdst = {
        .get_optmin     = SO_ORIGINAL_DST,
        .get_optmax     = SO_ORIGINAL_DST+1,
        .get            = &getorigdst,
+       .owner          = THIS_MODULE,
 };
 
-struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 = {
+struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 __read_mostly = {
        .l3proto         = PF_INET,
        .name            = "ipv4",
        .pkt_to_tuple    = ipv4_pkt_to_tuple,
        .invert_tuple    = ipv4_invert_tuple,
        .print_tuple     = ipv4_print_tuple,
-       .print_conntrack = ipv4_print_conntrack,
-       .prepare         = ipv4_prepare,
-       .get_features    = ipv4_get_features,
-#if defined(CONFIG_NF_CT_NETLINK) || \
-    defined(CONFIG_NF_CT_NETLINK_MODULE)
-       .tuple_to_nfattr = ipv4_tuple_to_nfattr,
-       .nfattr_to_tuple = ipv4_nfattr_to_tuple,
+       .get_l4proto     = ipv4_get_l4proto,
+#if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
+       .tuple_to_nlattr = ipv4_tuple_to_nlattr,
+       .nlattr_to_tuple = ipv4_nlattr_to_tuple,
+       .nla_policy      = ipv4_nla_policy,
+#endif
+#if defined(CONFIG_SYSCTL) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
+       .ctl_table_path  = nf_net_ipv4_netfilter_sysctl_path,
+       .ctl_table       = ip_ct_sysctl_table,
 #endif
        .me              = THIS_MODULE,
 };
 
-extern struct nf_conntrack_protocol nf_conntrack_protocol_tcp4;
-extern struct nf_conntrack_protocol nf_conntrack_protocol_udp4;
-extern struct nf_conntrack_protocol nf_conntrack_protocol_icmp;
-static int init_or_cleanup(int init)
+module_param_call(hashsize, nf_conntrack_set_hashsize, param_get_uint,
+                 &nf_conntrack_htable_size, 0600);
+
+MODULE_ALIAS("nf_conntrack-" __stringify(AF_INET));
+MODULE_ALIAS("ip_conntrack");
+MODULE_LICENSE("GPL");
+
+static int __init nf_conntrack_l3proto_ipv4_init(void)
 {
        int ret = 0;
 
-       if (!init) goto cleanup;
+       need_conntrack();
 
        ret = nf_register_sockopt(&so_getorigdst);
        if (ret < 0) {
                printk(KERN_ERR "Unable to register netfilter socket option\n");
-               goto cleanup_nothing;
+               return ret;
        }
 
-       ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_tcp4);
+       ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_tcp4);
        if (ret < 0) {
                printk("nf_conntrack_ipv4: can't register tcp.\n");
                goto cleanup_sockopt;
        }
 
-       ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_udp4);
+       ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_udp4);
        if (ret < 0) {
                printk("nf_conntrack_ipv4: can't register udp.\n");
                goto cleanup_tcp;
        }
 
-       ret = nf_conntrack_protocol_register(&nf_conntrack_protocol_icmp);
+       ret = nf_conntrack_l4proto_register(&nf_conntrack_l4proto_icmp);
        if (ret < 0) {
                printk("nf_conntrack_ipv4: can't register icmp.\n");
                goto cleanup_udp;
@@ -474,122 +453,54 @@ static int init_or_cleanup(int init)
                goto cleanup_icmp;
        }
 
-       ret = nf_register_hook(&ipv4_conntrack_defrag_ops);
+       ret = nf_register_hooks(ipv4_conntrack_ops,
+                               ARRAY_SIZE(ipv4_conntrack_ops));
        if (ret < 0) {
-               printk("nf_conntrack_ipv4: can't register pre-routing defrag hook.\n");
+               printk("nf_conntrack_ipv4: can't register hooks.\n");
                goto cleanup_ipv4;
        }
-       ret = nf_register_hook(&ipv4_conntrack_defrag_local_out_ops);
-       if (ret < 0) {
-               printk("nf_conntrack_ipv4: can't register local_out defrag hook.\n");
-               goto cleanup_defragops;
-       }
-
-       ret = nf_register_hook(&ipv4_conntrack_in_ops);
-       if (ret < 0) {
-               printk("nf_conntrack_ipv4: can't register pre-routing hook.\n");
-               goto cleanup_defraglocalops;
-       }
-
-       ret = nf_register_hook(&ipv4_conntrack_local_out_ops);
-       if (ret < 0) {
-               printk("nf_conntrack_ipv4: can't register local out hook.\n");
-               goto cleanup_inops;
-       }
-
-       ret = nf_register_hook(&ipv4_conntrack_helper_in_ops);
-       if (ret < 0) {
-               printk("nf_conntrack_ipv4: can't register local helper hook.\n");
-               goto cleanup_inandlocalops;
-       }
-
-       ret = nf_register_hook(&ipv4_conntrack_helper_out_ops);
-       if (ret < 0) {
-               printk("nf_conntrack_ipv4: can't register postrouting helper hook.\n");
-               goto cleanup_helperinops;
-       }
-
-       ret = nf_register_hook(&ipv4_conntrack_out_ops);
-       if (ret < 0) {
-               printk("nf_conntrack_ipv4: can't register post-routing hook.\n");
-               goto cleanup_helperoutops;
-       }
-
-       ret = nf_register_hook(&ipv4_conntrack_local_in_ops);
-       if (ret < 0) {
-               printk("nf_conntrack_ipv4: can't register local in hook.\n");
-               goto cleanup_inoutandlocalops;
-       }
-
-#ifdef CONFIG_SYSCTL
-       nf_ct_ipv4_sysctl_header = register_sysctl_table(nf_ct_net_table, 0);
-       if (nf_ct_ipv4_sysctl_header == NULL) {
-               printk("nf_conntrack: can't register to sysctl.\n");
-               ret = -ENOMEM;
-               goto cleanup_localinops;
-       }
+#if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
+       ret = nf_conntrack_ipv4_compat_init();
+       if (ret < 0)
+               goto cleanup_hooks;
 #endif
-
-       /* For use by REJECT target */
-       ip_ct_attach = __nf_conntrack_attach;
-
        return ret;
-
- cleanup:
-       synchronize_net();
-       ip_ct_attach = NULL;
-#ifdef CONFIG_SYSCTL
-       unregister_sysctl_table(nf_ct_ipv4_sysctl_header);
- cleanup_localinops:
+#if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
+ cleanup_hooks:
+       nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
 #endif
-       nf_unregister_hook(&ipv4_conntrack_local_in_ops);
- cleanup_inoutandlocalops:
-       nf_unregister_hook(&ipv4_conntrack_out_ops);
- cleanup_helperoutops:
-       nf_unregister_hook(&ipv4_conntrack_helper_out_ops);
- cleanup_helperinops:
-       nf_unregister_hook(&ipv4_conntrack_helper_in_ops);
- cleanup_inandlocalops:
-       nf_unregister_hook(&ipv4_conntrack_local_out_ops);
- cleanup_inops:
-       nf_unregister_hook(&ipv4_conntrack_in_ops);
- cleanup_defraglocalops:
-       nf_unregister_hook(&ipv4_conntrack_defrag_local_out_ops);
- cleanup_defragops:
-       nf_unregister_hook(&ipv4_conntrack_defrag_ops);
  cleanup_ipv4:
        nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
  cleanup_icmp:
-       nf_conntrack_protocol_unregister(&nf_conntrack_protocol_icmp);
+       nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
  cleanup_udp:
-       nf_conntrack_protocol_unregister(&nf_conntrack_protocol_udp4);
+       nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
  cleanup_tcp:
-       nf_conntrack_protocol_unregister(&nf_conntrack_protocol_tcp4);
+       nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
  cleanup_sockopt:
        nf_unregister_sockopt(&so_getorigdst);
- cleanup_nothing:
        return ret;
 }
 
-MODULE_LICENSE("GPL");
-
-static int __init init(void)
-{
-       need_nf_conntrack();
-       return init_or_cleanup(1);
-}
-
-static void __exit fini(void)
+static void __exit nf_conntrack_l3proto_ipv4_fini(void)
 {
-       init_or_cleanup(0);
+       synchronize_net();
+#if defined(CONFIG_PROC_FS) && defined(CONFIG_NF_CONNTRACK_PROC_COMPAT)
+       nf_conntrack_ipv4_compat_fini();
+#endif
+       nf_unregister_hooks(ipv4_conntrack_ops, ARRAY_SIZE(ipv4_conntrack_ops));
+       nf_conntrack_l3proto_unregister(&nf_conntrack_l3proto_ipv4);
+       nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_icmp);
+       nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_udp4);
+       nf_conntrack_l4proto_unregister(&nf_conntrack_l4proto_tcp4);
+       nf_unregister_sockopt(&so_getorigdst);
 }
 
-module_init(init);
-module_exit(fini);
+module_init(nf_conntrack_l3proto_ipv4_init);
+module_exit(nf_conntrack_l3proto_ipv4_fini);
 
-void need_ip_conntrack(void)
+void need_ipv4_conntrack(void)
 {
+       return;
 }
-
-EXPORT_SYMBOL(need_ip_conntrack);
-EXPORT_SYMBOL(nf_ct_ipv4_gather_frags);
+EXPORT_SYMBOL_GPL(need_ipv4_conntrack);