netfilter: nf_conntrack: support conntrack templates
[safe/jmp/linux-2.6] / net / ipv6 / netfilter / nf_conntrack_l3proto_ipv6.c
index dc442fb..55ce22e 100644 (file)
 #include <linux/icmp.h>
 #include <linux/sysctl.h>
 #include <net/ipv6.h>
+#include <net/inet_frag.h>
 
+#include <linux/netfilter_bridge.h>
 #include <linux/netfilter_ipv6.h>
 #include <net/netfilter/nf_conntrack.h>
 #include <net/netfilter/nf_conntrack_helper.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/ipv6/nf_conntrack_ipv6.h>
+#include <net/netfilter/nf_log.h>
 
-#if 0
-#define DEBUGP printk
-#else
-#define DEBUGP(format, args...)
-#endif
-
-static int ipv6_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
-                            struct nf_conntrack_tuple *tuple)
+static bool ipv6_pkt_to_tuple(const struct sk_buff *skb, unsigned int nhoff,
+                             struct nf_conntrack_tuple *tuple)
 {
-       u_int32_t _addrs[8], *ap;
+       const u_int32_t *ap;
+       u_int32_t _addrs[8];
 
        ap = skb_header_pointer(skb, nhoff + offsetof(struct ipv6hdr, saddr),
                                sizeof(_addrs), _addrs);
        if (ap == NULL)
-               return 0;
+               return false;
 
        memcpy(tuple->src.u3.ip6, ap, sizeof(tuple->src.u3.ip6));
        memcpy(tuple->dst.u3.ip6, ap + 4, sizeof(tuple->dst.u3.ip6));
 
-       return 1;
+       return true;
 }
 
-static int ipv6_invert_tuple(struct nf_conntrack_tuple *tuple,
-                            const struct nf_conntrack_tuple *orig)
+static bool ipv6_invert_tuple(struct nf_conntrack_tuple *tuple,
+                             const struct nf_conntrack_tuple *orig)
 {
        memcpy(tuple->src.u3.ip6, orig->dst.u3.ip6, sizeof(tuple->src.u3.ip6));
        memcpy(tuple->dst.u3.ip6, orig->src.u3.ip6, sizeof(tuple->dst.u3.ip6));
 
-       return 1;
+       return true;
 }
 
 static int ipv6_print_tuple(struct seq_file *s,
                            const struct nf_conntrack_tuple *tuple)
 {
-       return seq_printf(s, "src=" NIP6_FMT " dst=" NIP6_FMT " ",
-                         NIP6(*((struct in6_addr *)tuple->src.u3.ip6)),
-                         NIP6(*((struct in6_addr *)tuple->dst.u3.ip6)));
-}
-
-static int ipv6_print_conntrack(struct seq_file *s,
-                               const struct nf_conn *conntrack)
-{
-       return 0;
+       return seq_printf(s, "src=%pI6 dst=%pI6 ",
+                         tuple->src.u3.ip6, tuple->dst.u3.ip6);
 }
 
 /*
@@ -92,8 +84,8 @@ static int ipv6_print_conntrack(struct seq_file *s,
  *        - Note also special handling of AUTH header. Thanks to IPsec wizards.
  */
 
-int nf_ct_ipv6_skip_exthdr(struct sk_buff *skb, int start, u8 *nexthdrp,
-                          int len)
+static int nf_ct_ipv6_skip_exthdr(const struct sk_buff *skb, int start,
+                                 u8 *nexthdrp, int len)
 {
        u8 nexthdr = *nexthdrp;
 
@@ -123,22 +115,25 @@ int nf_ct_ipv6_skip_exthdr(struct sk_buff *skb, int start, u8 *nexthdrp,
        return start;
 }
 
-static int
-ipv6_prepare(struct sk_buff **pskb, unsigned int hooknum, unsigned int *dataoff,
-            u_int8_t *protonum)
+static int ipv6_get_l4proto(const struct sk_buff *skb, unsigned int nhoff,
+                           unsigned int *dataoff, u_int8_t *protonum)
 {
-       unsigned int extoff = (u8 *)(ipv6_hdr(*pskb) + 1) - (*pskb)->data;
-       unsigned char pnum = ipv6_hdr(*pskb)->nexthdr;
-       int protoff = nf_ct_ipv6_skip_exthdr(*pskb, extoff, &pnum,
-                                            (*pskb)->len - extoff);
+       unsigned int extoff = nhoff + sizeof(struct ipv6hdr);
+       unsigned char pnum;
+       int protoff;
+
+       if (skb_copy_bits(skb, nhoff + offsetof(struct ipv6hdr, nexthdr),
+                         &pnum, sizeof(pnum)) != 0) {
+               pr_debug("ip6_conntrack_core: can't get nexthdr\n");
+               return -NF_ACCEPT;
+       }
+       protoff = nf_ct_ipv6_skip_exthdr(skb, extoff, &pnum, skb->len - extoff);
        /*
-        * (protoff == (*pskb)->len) mean that the packet doesn't have no data
+        * (protoff == skb->len) mean that the packet doesn't have no data
         * except of IPv6 & ext headers. but it's tracked anyway. - YK
         */
-       if ((protoff < 0) || (protoff > (*pskb)->len)) {
-               DEBUGP("ip6_conntrack_core: can't find proto in pkt\n");
-               NF_CT_STAT_INC_ATOMIC(error);
-               NF_CT_STAT_INC_ATOMIC(invalid);
+       if ((protoff < 0) || (protoff > skb->len)) {
+               pr_debug("ip6_conntrack_core: can't find proto in pkt\n");
                return -NF_ACCEPT;
        }
 
@@ -147,51 +142,69 @@ ipv6_prepare(struct sk_buff **pskb, unsigned int hooknum, unsigned int *dataoff,
        return NF_ACCEPT;
 }
 
-static u_int32_t ipv6_get_features(const struct nf_conntrack_tuple *tuple)
-{
-       return NF_CT_F_BASIC;
-}
-
 static unsigned int ipv6_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 *))
 {
        struct nf_conn *ct;
-       struct nf_conn_help *help;
+       const struct nf_conn_help *help;
+       const struct nf_conntrack_helper *helper;
        enum ip_conntrack_info ctinfo;
        unsigned int ret, protoff;
-       unsigned int extoff = (u8 *)(ipv6_hdr(*pskb) + 1) - (*pskb)->data;
-       unsigned char pnum = ipv6_hdr(*pskb)->nexthdr;
+       unsigned int extoff = (u8 *)(ipv6_hdr(skb) + 1) - skb->data;
+       unsigned char pnum = ipv6_hdr(skb)->nexthdr;
 
 
        /* 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)
                goto out;
 
        help = nfct_help(ct);
-       if (!help || !help->helper)
+       if (!help)
+               goto out;
+       /* rcu_read_lock()ed by nf_hook_slow */
+       helper = rcu_dereference(help->helper);
+       if (!helper)
                goto out;
 
-       protoff = nf_ct_ipv6_skip_exthdr(*pskb, extoff, &pnum,
-                                        (*pskb)->len - extoff);
-       if (protoff > (*pskb)->len || pnum == NEXTHDR_FRAGMENT) {
-               DEBUGP("proto header not found\n");
+       protoff = nf_ct_ipv6_skip_exthdr(skb, extoff, &pnum,
+                                        skb->len - extoff);
+       if (protoff > skb->len || pnum == NEXTHDR_FRAGMENT) {
+               pr_debug("proto header not found\n");
                return NF_ACCEPT;
        }
 
-       ret = help->helper->help(pskb, protoff, ct, ctinfo);
-       if (ret != NF_ACCEPT)
+       ret = helper->help(skb, protoff, ct, ctinfo);
+       if (ret != NF_ACCEPT) {
+               nf_log_packet(NFPROTO_IPV6, hooknum, skb, in, out, NULL,
+                             "nf_ct_%s: dropping packet", helper->name);
                return ret;
+       }
 out:
        /* We've seen it coming out the other side: confirm it */
-       return nf_conntrack_confirm(pskb);
+       return nf_conntrack_confirm(skb);
+}
+
+static enum ip6_defrag_users nf_ct6_defrag_user(unsigned int hooknum,
+                                               struct sk_buff *skb)
+{
+#ifdef CONFIG_BRIDGE_NETFILTER
+       if (skb->nf_bridge &&
+           skb->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)
+               return IP6_DEFRAG_CONNTRACK_BRIDGE_IN;
+#endif
+       if (hooknum == NF_INET_PRE_ROUTING)
+               return IP6_DEFRAG_CONNTRACK_IN;
+       else
+               return IP6_DEFRAG_CONNTRACK_OUT;
+
 }
 
 static unsigned int ipv6_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 *))
@@ -199,17 +212,16 @@ static unsigned int ipv6_defrag(unsigned int hooknum,
        struct sk_buff *reasm;
 
        /* Previously seen (loopback)?  */
-       if ((*pskb)->nfct)
+       if (skb->nfct && !nf_ct_is_template((struct nf_conn *)skb->nfct))
                return NF_ACCEPT;
 
-       reasm = nf_ct_frag6_gather(*pskb);
-
+       reasm = nf_ct_frag6_gather(skb, nf_ct6_defrag_user(hooknum, skb));
        /* queued */
        if (reasm == NULL)
                return NF_STOLEN;
 
        /* error occured or not fragmented */
-       if (reasm == *pskb)
+       if (reasm == skb)
                return NF_ACCEPT;
 
        nf_ct_frag6_output(hooknum, reasm, (struct net_device *)in,
@@ -218,13 +230,12 @@ static unsigned int ipv6_defrag(unsigned int hooknum,
        return NF_STOLEN;
 }
 
-static unsigned int ipv6_conntrack_in(unsigned int hooknum,
-                                     struct sk_buff **pskb,
-                                     const struct net_device *in,
-                                     const struct net_device *out,
-                                     int (*okfn)(struct sk_buff *))
+static unsigned int __ipv6_conntrack_in(struct net *net,
+                                       unsigned int hooknum,
+                                       struct sk_buff *skb,
+                                       int (*okfn)(struct sk_buff *))
 {
-       struct sk_buff *reasm = (*pskb)->nfct_reasm;
+       struct sk_buff *reasm = skb->nfct_reasm;
 
        /* This packet is fragmented and has reassembled packet. */
        if (reasm) {
@@ -232,167 +243,148 @@ static unsigned int ipv6_conntrack_in(unsigned int hooknum,
                if (!reasm->nfct) {
                        unsigned int ret;
 
-                       ret = nf_conntrack_in(PF_INET6, hooknum, &reasm);
+                       ret = nf_conntrack_in(net, PF_INET6, hooknum, reasm);
                        if (ret != NF_ACCEPT)
                                return ret;
                }
                nf_conntrack_get(reasm->nfct);
-               (*pskb)->nfct = reasm->nfct;
-               (*pskb)->nfctinfo = reasm->nfctinfo;
+               skb->nfct = reasm->nfct;
+               skb->nfctinfo = reasm->nfctinfo;
                return NF_ACCEPT;
        }
 
-       return nf_conntrack_in(PF_INET6, hooknum, pskb);
+       return nf_conntrack_in(net, PF_INET6, hooknum, skb);
+}
+
+static unsigned int ipv6_conntrack_in(unsigned int hooknum,
+                                     struct sk_buff *skb,
+                                     const struct net_device *in,
+                                     const struct net_device *out,
+                                     int (*okfn)(struct sk_buff *))
+{
+       return __ipv6_conntrack_in(dev_net(in), hooknum, skb, okfn);
 }
 
 static unsigned int ipv6_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 ipv6hdr)) {
+       if (skb->len < sizeof(struct ipv6hdr)) {
                if (net_ratelimit())
                        printk("ipv6_conntrack_local: packet too short\n");
                return NF_ACCEPT;
        }
-       return ipv6_conntrack_in(hooknum, pskb, in, out, okfn);
+       return __ipv6_conntrack_in(dev_net(out), hooknum, skb, okfn);
 }
 
-static struct nf_hook_ops ipv6_conntrack_ops[] = {
+static struct nf_hook_ops ipv6_conntrack_ops[] __read_mostly = {
        {
                .hook           = ipv6_defrag,
                .owner          = THIS_MODULE,
-               .pf             = PF_INET6,
-               .hooknum        = NF_IP6_PRE_ROUTING,
+               .pf             = NFPROTO_IPV6,
+               .hooknum        = NF_INET_PRE_ROUTING,
                .priority       = NF_IP6_PRI_CONNTRACK_DEFRAG,
        },
        {
                .hook           = ipv6_conntrack_in,
                .owner          = THIS_MODULE,
-               .pf             = PF_INET6,
-               .hooknum        = NF_IP6_PRE_ROUTING,
+               .pf             = NFPROTO_IPV6,
+               .hooknum        = NF_INET_PRE_ROUTING,
                .priority       = NF_IP6_PRI_CONNTRACK,
        },
        {
                .hook           = ipv6_conntrack_local,
                .owner          = THIS_MODULE,
-               .pf             = PF_INET6,
-               .hooknum        = NF_IP6_LOCAL_OUT,
+               .pf             = NFPROTO_IPV6,
+               .hooknum        = NF_INET_LOCAL_OUT,
                .priority       = NF_IP6_PRI_CONNTRACK,
        },
        {
                .hook           = ipv6_defrag,
                .owner          = THIS_MODULE,
-               .pf             = PF_INET6,
-               .hooknum        = NF_IP6_LOCAL_OUT,
+               .pf             = NFPROTO_IPV6,
+               .hooknum        = NF_INET_LOCAL_OUT,
                .priority       = NF_IP6_PRI_CONNTRACK_DEFRAG,
        },
        {
                .hook           = ipv6_confirm,
                .owner          = THIS_MODULE,
-               .pf             = PF_INET6,
-               .hooknum        = NF_IP6_POST_ROUTING,
+               .pf             = NFPROTO_IPV6,
+               .hooknum        = NF_INET_POST_ROUTING,
                .priority       = NF_IP6_PRI_LAST,
        },
        {
                .hook           = ipv6_confirm,
                .owner          = THIS_MODULE,
-               .pf             = PF_INET6,
-               .hooknum        = NF_IP6_LOCAL_IN,
+               .pf             = NFPROTO_IPV6,
+               .hooknum        = NF_INET_LOCAL_IN,
                .priority       = NF_IP6_PRI_LAST-1,
        },
 };
 
-#ifdef CONFIG_SYSCTL
-static ctl_table nf_ct_ipv6_sysctl_table[] = {
-       {
-               .ctl_name       = NET_NF_CONNTRACK_FRAG6_TIMEOUT,
-               .procname       = "nf_conntrack_frag6_timeout",
-               .data           = &nf_ct_frag6_timeout,
-               .maxlen         = sizeof(unsigned int),
-               .mode           = 0644,
-               .proc_handler   = &proc_dointvec_jiffies,
-       },
-       {
-               .ctl_name       = NET_NF_CONNTRACK_FRAG6_LOW_THRESH,
-               .procname       = "nf_conntrack_frag6_low_thresh",
-               .data           = &nf_ct_frag6_low_thresh,
-               .maxlen         = sizeof(unsigned int),
-               .mode           = 0644,
-               .proc_handler   = &proc_dointvec,
-       },
-       {
-               .ctl_name       = NET_NF_CONNTRACK_FRAG6_HIGH_THRESH,
-               .procname       = "nf_conntrack_frag6_high_thresh",
-               .data           = &nf_ct_frag6_high_thresh,
-               .maxlen         = sizeof(unsigned int),
-               .mode           = 0644,
-               .proc_handler   = &proc_dointvec,
-       },
-       { .ctl_name = 0 }
-};
-#endif
-
 #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 ipv6_tuple_to_nfattr(struct sk_buff *skb,
+static int ipv6_tuple_to_nlattr(struct sk_buff *skb,
                                const struct nf_conntrack_tuple *tuple)
 {
-       NFA_PUT(skb, CTA_IP_V6_SRC, sizeof(u_int32_t) * 4,
+       NLA_PUT(skb, CTA_IP_V6_SRC, sizeof(u_int32_t) * 4,
                &tuple->src.u3.ip6);
-       NFA_PUT(skb, CTA_IP_V6_DST, sizeof(u_int32_t) * 4,
+       NLA_PUT(skb, CTA_IP_V6_DST, sizeof(u_int32_t) * 4,
                &tuple->dst.u3.ip6);
        return 0;
 
-nfattr_failure:
+nla_put_failure:
        return -1;
 }
 
-static const size_t cta_min_ip[CTA_IP_MAX] = {
-       [CTA_IP_V6_SRC-1]       = sizeof(u_int32_t)*4,
-       [CTA_IP_V6_DST-1]       = sizeof(u_int32_t)*4,
+static const struct nla_policy ipv6_nla_policy[CTA_IP_MAX+1] = {
+       [CTA_IP_V6_SRC] = { .len = sizeof(u_int32_t)*4 },
+       [CTA_IP_V6_DST] = { .len = sizeof(u_int32_t)*4 },
 };
 
-static int ipv6_nfattr_to_tuple(struct nfattr *tb[],
+static int ipv6_nlattr_to_tuple(struct nlattr *tb[],
                                struct nf_conntrack_tuple *t)
 {
-       if (!tb[CTA_IP_V6_SRC-1] || !tb[CTA_IP_V6_DST-1])
+       if (!tb[CTA_IP_V6_SRC] || !tb[CTA_IP_V6_DST])
                return -EINVAL;
 
-       if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
-               return -EINVAL;
-
-       memcpy(&t->src.u3.ip6, NFA_DATA(tb[CTA_IP_V6_SRC-1]),
+       memcpy(&t->src.u3.ip6, nla_data(tb[CTA_IP_V6_SRC]),
               sizeof(u_int32_t) * 4);
-       memcpy(&t->dst.u3.ip6, NFA_DATA(tb[CTA_IP_V6_DST-1]),
+       memcpy(&t->dst.u3.ip6, nla_data(tb[CTA_IP_V6_DST]),
               sizeof(u_int32_t) * 4);
 
        return 0;
 }
+
+static int ipv6_nlattr_tuple_size(void)
+{
+       return nla_policy_len(ipv6_nla_policy, CTA_IP_MAX + 1);
+}
 #endif
 
-struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv6 = {
+struct nf_conntrack_l3proto nf_conntrack_l3proto_ipv6 __read_mostly = {
        .l3proto                = PF_INET6,
        .name                   = "ipv6",
        .pkt_to_tuple           = ipv6_pkt_to_tuple,
        .invert_tuple           = ipv6_invert_tuple,
        .print_tuple            = ipv6_print_tuple,
-       .print_conntrack        = ipv6_print_conntrack,
-       .prepare                = ipv6_prepare,
+       .get_l4proto            = ipv6_get_l4proto,
 #if defined(CONFIG_NF_CT_NETLINK) || defined(CONFIG_NF_CT_NETLINK_MODULE)
-       .tuple_to_nfattr        = ipv6_tuple_to_nfattr,
-       .nfattr_to_tuple        = ipv6_nfattr_to_tuple,
+       .tuple_to_nlattr        = ipv6_tuple_to_nlattr,
+       .nlattr_tuple_size      = ipv6_nlattr_tuple_size,
+       .nlattr_to_tuple        = ipv6_nlattr_to_tuple,
+       .nla_policy             = ipv6_nla_policy,
 #endif
 #ifdef CONFIG_SYSCTL
        .ctl_table_path         = nf_net_netfilter_sysctl_path,
        .ctl_table              = nf_ct_ipv6_sysctl_table,
 #endif
-       .get_features           = ipv6_get_features,
        .me                     = THIS_MODULE,
 };