net: skb->dst accessors
[safe/jmp/linux-2.6] / net / bridge / br_netfilter.c
index 141f069..d22f611 100644 (file)
@@ -58,11 +58,11 @@ static struct ctl_table_header *brnf_sysctl_header;
 static int brnf_call_iptables __read_mostly = 1;
 static int brnf_call_ip6tables __read_mostly = 1;
 static int brnf_call_arptables __read_mostly = 1;
-static int brnf_filter_vlan_tagged __read_mostly = 1;
-static int brnf_filter_pppoe_tagged __read_mostly = 1;
+static int brnf_filter_vlan_tagged __read_mostly = 0;
+static int brnf_filter_pppoe_tagged __read_mostly = 0;
 #else
-#define brnf_filter_vlan_tagged 1
-#define brnf_filter_pppoe_tagged 1
+#define brnf_filter_vlan_tagged 0
+#define brnf_filter_pppoe_tagged 0
 #endif
 
 static inline __be16 vlan_proto(const struct sk_buff *skb)
@@ -101,30 +101,42 @@ static inline __be16 pppoe_proto(const struct sk_buff *skb)
         pppoe_proto(skb) == htons(PPP_IPV6) && \
         brnf_filter_pppoe_tagged)
 
-/* We need these fake structures to make netfilter happy --
- * lots of places assume that skb->dst != NULL, which isn't
- * all that unreasonable.
- *
+static void fake_update_pmtu(struct dst_entry *dst, u32 mtu)
+{
+}
+
+static struct dst_ops fake_dst_ops = {
+       .family =               AF_INET,
+       .protocol =             cpu_to_be16(ETH_P_IP),
+       .update_pmtu =          fake_update_pmtu,
+       .entries =              ATOMIC_INIT(0),
+};
+
+/*
+ * Initialize bogus route table used to keep netfilter happy.
  * Currently, we fill in the PMTU entry because netfilter
  * refragmentation needs it, and the rt_flags entry because
  * ipt_REJECT needs it.  Future netfilter modules might
- * require us to fill additional fields. */
-static struct net_device __fake_net_device = {
-       .hard_header_len        = ETH_HLEN
-};
+ * require us to fill additional fields.
+ */
+void br_netfilter_rtable_init(struct net_bridge *br)
+{
+       struct rtable *rt = &br->fake_rtable;
+
+       atomic_set(&rt->u.dst.__refcnt, 1);
+       rt->u.dst.dev = br->dev;
+       rt->u.dst.path = &rt->u.dst;
+       rt->u.dst.metrics[RTAX_MTU - 1] = 1500;
+       rt->u.dst.flags = DST_NOXFRM;
+       rt->u.dst.ops = &fake_dst_ops;
+}
 
-static struct rtable __fake_rtable = {
-       .u = {
-               .dst = {
-                       .__refcnt               = ATOMIC_INIT(1),
-                       .dev                    = &__fake_net_device,
-                       .path                   = &__fake_rtable.u.dst,
-                       .metrics                = {[RTAX_MTU - 1] = 1500},
-                       .flags                  = DST_NOXFRM,
-               }
-       },
-       .rt_flags       = 0,
-};
+static inline struct rtable *bridge_parent_rtable(const struct net_device *dev)
+{
+       struct net_bridge_port *port = rcu_dereference(dev->br_port);
+
+       return port ? &port->br->fake_rtable : NULL;
+}
 
 static inline struct net_device *bridge_parent(const struct net_device *dev)
 {
@@ -216,6 +228,7 @@ int nf_bridge_copy_header(struct sk_buff *skb)
 static int br_nf_pre_routing_finish_ipv6(struct sk_buff *skb)
 {
        struct nf_bridge_info *nf_bridge = skb->nf_bridge;
+       struct rtable *rt;
 
        if (nf_bridge->mask & BRNF_PKT_TYPE) {
                skb->pkt_type = PACKET_OTHERHOST;
@@ -223,8 +236,13 @@ static int br_nf_pre_routing_finish_ipv6(struct sk_buff *skb)
        }
        nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
 
-       skb->dst = (struct dst_entry *)&__fake_rtable;
-       dst_hold(skb->dst);
+       rt = bridge_parent_rtable(nf_bridge->physindev);
+       if (!rt) {
+               kfree_skb(skb);
+               return 0;
+       }
+       dst_hold(&rt->u.dst);
+       skb_dst_set(skb, &rt->u.dst);
 
        skb->dev = nf_bridge->physindev;
        nf_bridge_push_encap_header(skb);
@@ -304,7 +322,7 @@ static int br_nf_pre_routing_finish_bridge(struct sk_buff *skb)
 
        skb->dev = bridge_parent(skb->dev);
        if (skb->dev) {
-               struct dst_entry *dst = skb->dst;
+               struct dst_entry *dst = skb_dst(skb);
 
                nf_bridge_pull_encap_header(skb);
 
@@ -322,6 +340,7 @@ static int br_nf_pre_routing_finish(struct sk_buff *skb)
        struct net_device *dev = skb->dev;
        struct iphdr *iph = ip_hdr(skb);
        struct nf_bridge_info *nf_bridge = skb->nf_bridge;
+       struct rtable *rt;
        int err;
 
        if (nf_bridge->mask & BRNF_PKT_TYPE) {
@@ -331,7 +350,6 @@ static int br_nf_pre_routing_finish(struct sk_buff *skb)
        nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
        if (dnat_took_place(skb)) {
                if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
-                       struct rtable *rt;
                        struct flowi fl = {
                                .nl_u = {
                                        .ip4_u = {
@@ -353,11 +371,11 @@ static int br_nf_pre_routing_finish(struct sk_buff *skb)
                        if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
                                goto free_skb;
 
-                       if (!ip_route_output_key(&rt, &fl)) {
+                       if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
                                /* - Bridged-and-DNAT'ed traffic doesn't
                                 *   require ip_forwarding. */
                                if (((struct dst_entry *)rt)->dev == dev) {
-                                       skb->dst = (struct dst_entry *)rt;
+                                       skb_dst_set(skb, (struct dst_entry *)rt);
                                        goto bridged_dnat;
                                }
                                /* we are sure that forwarding is disabled, so printing
@@ -371,7 +389,7 @@ free_skb:
                        kfree_skb(skb);
                        return 0;
                } else {
-                       if (skb->dst->dev == dev) {
+                       if (skb_dst(skb)->dev == dev) {
 bridged_dnat:
                                /* Tell br_nf_local_out this is a
                                 * bridged frame */
@@ -388,8 +406,13 @@ bridged_dnat:
                        skb->pkt_type = PACKET_HOST;
                }
        } else {
-               skb->dst = (struct dst_entry *)&__fake_rtable;
-               dst_hold(skb->dst);
+               rt = bridge_parent_rtable(nf_bridge->physindev);
+               if (!rt) {
+                       kfree_skb(skb);
+                       return 0;
+               }
+               dst_hold(&rt->u.dst);
+               skb_dst_set(skb, &rt->u.dst);
        }
 
        skb->dev = nf_bridge->physindev;
@@ -608,10 +631,10 @@ static unsigned int br_nf_local_in(unsigned int hook, struct sk_buff *skb,
                                   const struct net_device *out,
                                   int (*okfn)(struct sk_buff *))
 {
-       if (skb->dst == (struct dst_entry *)&__fake_rtable) {
-               dst_release(skb->dst);
-               skb->dst = NULL;
-       }
+       struct rtable *rt = skb_rtable(skb);
+
+       if (rt && rt == bridge_parent_rtable(in))
+               skb_dst_drop(skb);
 
        return NF_ACCEPT;
 }
@@ -649,7 +672,7 @@ static unsigned int br_nf_forward_ip(unsigned int hook, struct sk_buff *skb,
 {
        struct nf_bridge_info *nf_bridge;
        struct net_device *parent;
-       int pf;
+       u_int8_t pf;
 
        if (!skb->nf_bridge)
                return NF_ACCEPT;
@@ -666,8 +689,11 @@ static unsigned int br_nf_forward_ip(unsigned int hook, struct sk_buff *skb,
        if (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb) ||
            IS_PPPOE_IP(skb))
                pf = PF_INET;
-       else
+       else if (skb->protocol == htons(ETH_P_IPV6) || IS_VLAN_IPV6(skb) ||
+                IS_PPPOE_IPV6(skb))
                pf = PF_INET6;
+       else
+               return NF_ACCEPT;
 
        nf_bridge_pull_encap_header(skb);
 
@@ -711,7 +737,7 @@ static unsigned int br_nf_forward_arp(unsigned int hook, struct sk_buff *skb,
                return NF_ACCEPT;
        }
        *d = (struct net_device *)in;
-       NF_HOOK(NF_ARP, NF_ARP_FORWARD, skb, (struct net_device *)in,
+       NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, skb, (struct net_device *)in,
                (struct net_device *)out, br_nf_forward_finish);
 
        return NF_STOLEN;
@@ -765,15 +791,23 @@ static unsigned int br_nf_local_out(unsigned int hook, struct sk_buff *skb,
        return NF_STOLEN;
 }
 
+#if defined(CONFIG_NF_CONNTRACK_IPV4) || defined(CONFIG_NF_CONNTRACK_IPV4_MODULE)
 static int br_nf_dev_queue_xmit(struct sk_buff *skb)
 {
-       if (skb->protocol == htons(ETH_P_IP) &&
+       if (skb->nfct != NULL &&
+           (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb)) &&
            skb->len > skb->dev->mtu &&
            !skb_is_gso(skb))
                return ip_fragment(skb, br_dev_queue_push_xmit);
        else
                return br_dev_queue_push_xmit(skb);
 }
+#else
+static int br_nf_dev_queue_xmit(struct sk_buff *skb)
+{
+        return br_dev_queue_push_xmit(skb);
+}
+#endif
 
 /* PF_BRIDGE/POST_ROUTING ********************************************/
 static unsigned int br_nf_post_routing(unsigned int hook, struct sk_buff *skb,
@@ -783,7 +817,7 @@ static unsigned int br_nf_post_routing(unsigned int hook, struct sk_buff *skb,
 {
        struct nf_bridge_info *nf_bridge = skb->nf_bridge;
        struct net_device *realoutdev = bridge_parent(skb->dev);
-       int pf;
+       u_int8_t pf;
 
 #ifdef CONFIG_NETFILTER_DEBUG
        /* Be very paranoid. This probably won't happen anymore, but let's
@@ -808,11 +842,14 @@ static unsigned int br_nf_post_routing(unsigned int hook, struct sk_buff *skb,
        if (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb) ||
            IS_PPPOE_IP(skb))
                pf = PF_INET;
-       else
+       else if (skb->protocol == htons(ETH_P_IPV6) || IS_VLAN_IPV6(skb) ||
+                IS_PPPOE_IPV6(skb))
                pf = PF_INET6;
+       else
+               return NF_ACCEPT;
 
 #ifdef CONFIG_NETFILTER_DEBUG
-       if (skb->dst == NULL) {
+       if (skb_dst(skb) == NULL) {
                printk(KERN_INFO "br_netfilter post_routing: skb->dst == NULL\n");
                goto print_error;
        }
@@ -828,10 +865,6 @@ static unsigned int br_nf_post_routing(unsigned int hook, struct sk_buff *skb,
        nf_bridge_pull_encap_header(skb);
        nf_bridge_save_header(skb);
 
-#if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
-       if (nf_bridge->netoutdev)
-               realoutdev = nf_bridge->netoutdev;
-#endif
        NF_HOOK(pf, NF_INET_POST_ROUTING, skb, NULL, realoutdev,
                br_nf_dev_queue_xmit);
 
@@ -934,35 +967,35 @@ static ctl_table brnf_table[] = {
                .data           = &brnf_call_arptables,
                .maxlen         = sizeof(int),
                .mode           = 0644,
-               .proc_handler   = &brnf_sysctl_call_tables,
+               .proc_handler   = brnf_sysctl_call_tables,
        },
        {
                .procname       = "bridge-nf-call-iptables",
                .data           = &brnf_call_iptables,
                .maxlen         = sizeof(int),
                .mode           = 0644,
-               .proc_handler   = &brnf_sysctl_call_tables,
+               .proc_handler   = brnf_sysctl_call_tables,
        },
        {
                .procname       = "bridge-nf-call-ip6tables",
                .data           = &brnf_call_ip6tables,
                .maxlen         = sizeof(int),
                .mode           = 0644,
-               .proc_handler   = &brnf_sysctl_call_tables,
+               .proc_handler   = brnf_sysctl_call_tables,
        },
        {
                .procname       = "bridge-nf-filter-vlan-tagged",
                .data           = &brnf_filter_vlan_tagged,
                .maxlen         = sizeof(int),
                .mode           = 0644,
-               .proc_handler   = &brnf_sysctl_call_tables,
+               .proc_handler   = brnf_sysctl_call_tables,
        },
        {
                .procname       = "bridge-nf-filter-pppoe-tagged",
                .data           = &brnf_filter_pppoe_tagged,
                .maxlen         = sizeof(int),
                .mode           = 0644,
-               .proc_handler   = &brnf_sysctl_call_tables,
+               .proc_handler   = brnf_sysctl_call_tables,
        },
        { .ctl_name = 0 }
 };