[NETFILTER]: Introduce NF_INET_ hook values
[safe/jmp/linux-2.6] / net / ipv4 / netfilter / nf_conntrack_l3proto_ipv4.c
index 0654eaa..c91725a 100644 (file)
 #include <net/netfilter/nf_conntrack_core.h>
 #include <net/netfilter/ipv4/nf_conntrack_ipv4.h>
 
-#if 0
-#define DEBUGP printk
-#else
-#define DEBUGP(format, args...)
-#endif
-
 static int ipv4_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
                             struct nf_conntrack_tuple *tuple)
 {
@@ -69,63 +63,54 @@ static int ipv4_print_conntrack(struct seq_file *s,
 }
 
 /* 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);
+       err = ip_defrag(skb, user);
        local_bh_enable();
 
-       if (skb)
+       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 (ip_hdr(*pskb)->frag_off & htons(IP_OFFSET)) {
-               if (net_ratelimit()) {
-                       printk(KERN_ERR "ipv4_prepare: Frag of proto %u (hook=%u)\n",
-                       ip_hdr(*pskb)->protocol, hooknum);
-               }
-               return -NF_DROP;
-       }
+       struct iphdr _iph, *iph;
 
-       *dataoff = skb_network_offset(*pskb) + ip_hdrlen(*pskb);
-       *protonum = ip_hdr(*pskb)->protocol;
-
-       return NF_ACCEPT;
-}
+       iph = skb_header_pointer(skb, nhoff, sizeof(_iph), &_iph);
+       if (iph == NULL)
+               return -NF_DROP;
 
-int nf_nat_module_is_loaded = 0;
-EXPORT_SYMBOL_GPL(nf_nat_module_is_loaded);
+       /* Conntrack defragments packets, we might still see fragments
+        * inside ICMP packets though. */
+       if (iph->frag_off & htons(IP_OFFSET))
+               return -NF_DROP;
 
-static u_int32_t ipv4_get_features(const struct nf_conntrack_tuple *tuple)
-{
-       if (nf_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);
+       return nf_conntrack_confirm(skb);
 }
 
 static unsigned int ipv4_conntrack_help(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 *))
@@ -133,69 +118,69 @@ static unsigned int ipv4_conntrack_help(unsigned int hooknum,
        struct nf_conn *ct;
        enum ip_conntrack_info ctinfo;
        struct nf_conn_help *help;
+       struct nf_conntrack_helper *helper;
 
        /* This is where we call the helper: as the packet goes out. */
-       ct = nf_ct_get(*pskb, &ctinfo);
+       ct = nf_ct_get(skb, &ctinfo);
        if (!ct || ctinfo == IP_CT_RELATED + IP_CT_IS_REPLY)
                return NF_ACCEPT;
 
        help = nfct_help(ct);
-       if (!help || !help->helper)
+       if (!help)
                return NF_ACCEPT;
-
-       return help->helper->help(pskb,
-                                 skb_network_offset(*pskb) + ip_hdrlen(*pskb),
-                                 ct, ctinfo);
+       /* rcu_read_lock()ed by nf_hook_slow */
+       helper = rcu_dereference(help->helper);
+       if (!helper)
+               return NF_ACCEPT;
+       return helper->help(skb, skb_network_offset(skb) + ip_hdrlen(skb),
+                           ct, ctinfo);
 }
 
 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 (ip_hdr(*pskb)->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(PF_INET, hooknum, skb);
 }
 
 static unsigned int ipv4_conntrack_local(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 *))
 {
        /* root is playing with raw sockets. */
-       if ((*pskb)->len < sizeof(struct iphdr)
-           || ip_hdrlen(*pskb) < 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(PF_INET, hooknum, skb);
 }
 
 /* Connection tracking may drop packets, but never alters them, so
@@ -205,56 +190,56 @@ static struct nf_hook_ops ipv4_conntrack_ops[] = {
                .hook           = ipv4_conntrack_defrag,
                .owner          = THIS_MODULE,
                .pf             = PF_INET,
-               .hooknum        = NF_IP_PRE_ROUTING,
+               .hooknum        = NF_INET_PRE_ROUTING,
                .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
        },
        {
                .hook           = ipv4_conntrack_in,
                .owner          = THIS_MODULE,
                .pf             = PF_INET,
-               .hooknum        = NF_IP_PRE_ROUTING,
+               .hooknum        = NF_INET_PRE_ROUTING,
                .priority       = NF_IP_PRI_CONNTRACK,
        },
        {
                .hook           = ipv4_conntrack_defrag,
                .owner          = THIS_MODULE,
                .pf             = PF_INET,
-               .hooknum        = NF_IP_LOCAL_OUT,
+               .hooknum        = NF_INET_LOCAL_OUT,
                .priority       = NF_IP_PRI_CONNTRACK_DEFRAG,
        },
        {
                .hook           = ipv4_conntrack_local,
                .owner          = THIS_MODULE,
                .pf             = PF_INET,
-               .hooknum        = NF_IP_LOCAL_OUT,
+               .hooknum        = NF_INET_LOCAL_OUT,
                .priority       = NF_IP_PRI_CONNTRACK,
        },
        {
                .hook           = ipv4_conntrack_help,
                .owner          = THIS_MODULE,
                .pf             = PF_INET,
-               .hooknum        = NF_IP_POST_ROUTING,
+               .hooknum        = NF_INET_POST_ROUTING,
                .priority       = NF_IP_PRI_CONNTRACK_HELPER,
        },
        {
                .hook           = ipv4_conntrack_help,
                .owner          = THIS_MODULE,
                .pf             = PF_INET,
-               .hooknum        = NF_IP_LOCAL_IN,
+               .hooknum        = NF_INET_LOCAL_IN,
                .priority       = NF_IP_PRI_CONNTRACK_HELPER,
        },
        {
                .hook           = ipv4_confirm,
                .owner          = THIS_MODULE,
                .pf             = PF_INET,
-               .hooknum        = NF_IP_POST_ROUTING,
+               .hooknum        = NF_INET_POST_ROUTING,
                .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
        },
        {
                .hook           = ipv4_confirm,
                .owner          = THIS_MODULE,
                .pf             = PF_INET,
-               .hooknum        = NF_IP_LOCAL_IN,
+               .hooknum        = NF_INET_LOCAL_IN,
                .priority       = NF_IP_PRI_CONNTRACK_CONFIRM,
        },
 };
@@ -334,17 +319,17 @@ 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(&tuple);
        if (h) {
                struct sockaddr_in sin;
                struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
@@ -356,17 +341,17 @@ getorigdst(struct sock *sk, int optval, void __user *user, int *len)
                        .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;
 }
 
@@ -375,35 +360,32 @@ getorigdst(struct sock *sk, int optval, void __user *user, int *len)
 #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),
+       NLA_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),
+       NLA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t),
                &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])
+       if (!tb[CTA_IP_V4_SRC] || !tb[CTA_IP_V4_DST])
                return -EINVAL;
 
-       if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
-               return -EINVAL;
-
-       t->src.u3.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
-       t->dst.u3.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
+       t->src.u3.ip = *(__be32 *)nla_data(tb[CTA_IP_V4_SRC]);
+       t->dst.u3.ip = *(__be32 *)nla_data(tb[CTA_IP_V4_DST]);
 
        return 0;
 }
@@ -414,20 +396,21 @@ 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,
+       .get_l4proto     = ipv4_get_l4proto,
 #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,
+       .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,
@@ -436,6 +419,9 @@ struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv4 = {
        .me              = THIS_MODULE,
 };
 
+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");
@@ -521,3 +507,9 @@ static void __exit nf_conntrack_l3proto_ipv4_fini(void)
 
 module_init(nf_conntrack_l3proto_ipv4_init);
 module_exit(nf_conntrack_l3proto_ipv4_fini);
+
+void need_ipv4_conntrack(void)
+{
+       return;
+}
+EXPORT_SYMBOL_GPL(need_ipv4_conntrack);