[NETFILTER]: Introduce NF_INET_ hook values
[safe/jmp/linux-2.6] / net / bridge / br_netfilter.c
index ac181be..f9ef3e5 100644 (file)
 #include <linux/if_arp.h>
 #include <linux/if_ether.h>
 #include <linux/if_vlan.h>
+#include <linux/if_pppox.h>
+#include <linux/ppp_defs.h>
 #include <linux/netfilter_bridge.h>
 #include <linux/netfilter_ipv4.h>
 #include <linux/netfilter_ipv6.h>
 #include <linux/netfilter_arp.h>
 #include <linux/in_route.h>
+#include <linux/inetdevice.h>
 
 #include <net/ip.h>
 #include <net/ipv6.h>
 #include <net/route.h>
 
 #include <asm/uaccess.h>
-#include <asm/checksum.h>
 #include "br_private.h"
 #ifdef CONFIG_SYSCTL
 #include <linux/sysctl.h>
@@ -48,8 +50,8 @@
 
 #define skb_origaddr(skb)       (((struct bridge_skb_cb *) \
                                 (skb->nf_bridge->data))->daddr.ipv4)
-#define store_orig_dstaddr(skb)         (skb_origaddr(skb) = (skb)->nh.iph->daddr)
-#define dnat_took_place(skb)    (skb_origaddr(skb) != (skb)->nh.iph->daddr)
+#define store_orig_dstaddr(skb)         (skb_origaddr(skb) = ip_hdr(skb)->daddr)
+#define dnat_took_place(skb)    (skb_origaddr(skb) != ip_hdr(skb)->daddr)
 
 #ifdef CONFIG_SYSCTL
 static struct ctl_table_header *brnf_sysctl_header;
@@ -57,21 +59,20 @@ 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;
 #else
 #define brnf_filter_vlan_tagged 1
+#define brnf_filter_pppoe_tagged 1
 #endif
 
-int brnf_deferred_hooks;
-EXPORT_SYMBOL_GPL(brnf_deferred_hooks);
-
-static __be16 inline vlan_proto(const struct sk_buff *skb)
+static inline __be16 vlan_proto(const struct sk_buff *skb)
 {
        return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
 }
 
 #define IS_VLAN_IP(skb) \
        (skb->protocol == htons(ETH_P_8021Q) && \
-        vlan_proto(skb) == htons(ETH_P_IP) &&  \
+        vlan_proto(skb) == htons(ETH_P_IP) &&  \
         brnf_filter_vlan_tagged)
 
 #define IS_VLAN_IPV6(skb) \
@@ -84,6 +85,22 @@ static __be16 inline vlan_proto(const struct sk_buff *skb)
         vlan_proto(skb) == htons(ETH_P_ARP) && \
         brnf_filter_vlan_tagged)
 
+static inline __be16 pppoe_proto(const struct sk_buff *skb)
+{
+       return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
+                           sizeof(struct pppoe_hdr)));
+}
+
+#define IS_PPPOE_IP(skb) \
+       (skb->protocol == htons(ETH_P_PPP_SES) && \
+        pppoe_proto(skb) == htons(PPP_IP) && \
+        brnf_filter_pppoe_tagged)
+
+#define IS_PPPOE_IPV6(skb) \
+       (skb->protocol == htons(ETH_P_PPP_SES) && \
+        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.
@@ -125,14 +142,53 @@ static inline struct nf_bridge_info *nf_bridge_alloc(struct sk_buff *skb)
        return skb->nf_bridge;
 }
 
-static inline void nf_bridge_save_header(struct sk_buff *skb)
+static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
 {
-        int header_size = ETH_HLEN;
+       struct nf_bridge_info *nf_bridge = skb->nf_bridge;
 
-       if (skb->protocol == htons(ETH_P_8021Q))
-               header_size += VLAN_HLEN;
+       if (atomic_read(&nf_bridge->use) > 1) {
+               struct nf_bridge_info *tmp = nf_bridge_alloc(skb);
 
-       memcpy(skb->nf_bridge->data, skb->data - header_size, header_size);
+               if (tmp) {
+                       memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info));
+                       atomic_set(&tmp->use, 1);
+                       nf_bridge_put(nf_bridge);
+               }
+               nf_bridge = tmp;
+       }
+       return nf_bridge;
+}
+
+static inline void nf_bridge_push_encap_header(struct sk_buff *skb)
+{
+       unsigned int len = nf_bridge_encap_header_len(skb);
+
+       skb_push(skb, len);
+       skb->network_header -= len;
+}
+
+static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
+{
+       unsigned int len = nf_bridge_encap_header_len(skb);
+
+       skb_pull(skb, len);
+       skb->network_header += len;
+}
+
+static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
+{
+       unsigned int len = nf_bridge_encap_header_len(skb);
+
+       skb_pull_rcsum(skb, len);
+       skb->network_header += len;
+}
+
+static inline void nf_bridge_save_header(struct sk_buff *skb)
+{
+       int header_size = ETH_HLEN + nf_bridge_encap_header_len(skb);
+
+       skb_copy_from_linear_data_offset(skb, -header_size,
+                                        skb->nf_bridge->data, header_size);
 }
 
 /*
@@ -142,19 +198,15 @@ static inline void nf_bridge_save_header(struct sk_buff *skb)
 int nf_bridge_copy_header(struct sk_buff *skb)
 {
        int err;
-        int header_size = ETH_HLEN;
-
-       if (skb->protocol == htons(ETH_P_8021Q))
-               header_size += VLAN_HLEN;
+       int header_size = ETH_HLEN + nf_bridge_encap_header_len(skb);
 
-       err = skb_cow(skb, header_size);
+       err = skb_cow_head(skb, header_size);
        if (err)
                return err;
 
-       memcpy(skb->data - header_size, skb->nf_bridge->data, header_size);
-
-       if (skb->protocol == htons(ETH_P_8021Q))
-               __skb_push(skb, VLAN_HLEN);
+       skb_copy_to_linear_data_offset(skb, -header_size,
+                                      skb->nf_bridge->data, header_size);
+       __skb_push(skb, nf_bridge_encap_header_len(skb));
        return 0;
 }
 
@@ -175,10 +227,7 @@ static int br_nf_pre_routing_finish_ipv6(struct sk_buff *skb)
        dst_hold(skb->dst);
 
        skb->dev = nf_bridge->physindev;
-       if (skb->protocol == htons(ETH_P_8021Q)) {
-               skb_push(skb, VLAN_HLEN);
-               skb->nh.raw -= VLAN_HLEN;
-       }
+       nf_bridge_push_encap_header(skb);
        NF_HOOK_THRESH(PF_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
                       br_handle_frame_finish, 1);
 
@@ -215,17 +264,22 @@ static void __br_dnat_complain(void)
  * Let us first consider the case that ip_route_input() succeeds:
  *
  * If skb->dst->dev equals the logical bridge device the packet
- * came in on, we can consider this bridging. We then call
- * skb->dst->output() which will make the packet enter br_nf_local_out()
+ * came in on, we can consider this bridging. The packet is passed
+ * through the neighbour output function to build a new destination
+ * MAC address, which will make the packet enter br_nf_local_out()
  * not much later. In that function it is assured that the iptables
  * FORWARD chain is traversed for the packet.
  *
  * Otherwise, the packet is considered to be routed and we just
  * change the destination MAC address so that the packet will
- * later be passed up to the IP stack to be routed.
+ * later be passed up to the IP stack to be routed. For a redirected
+ * packet, ip_route_input() will give back the localhost as output device,
+ * which differs from the bridge device.
  *
  * Let us now consider the case that ip_route_input() fails:
  *
+ * This can be because the destination address is martian, in which case
+ * the packet will be dropped.
  * After a "echo '0' > /proc/sys/net/ipv4/ip_forward" ip_route_input()
  * will fail, while __ip_route_output_key() will return success. The source
  * address for __ip_route_output_key() is set to zero, so __ip_route_output_key
@@ -238,7 +292,8 @@ static void __br_dnat_complain(void)
  *
  * --Lennert, 20020411
  * --Bart, 20020416 (updated)
- * --Bart, 20021007 (updated) */
+ * --Bart, 20021007 (updated)
+ * --Bart, 20062711 (updated) */
 static int br_nf_pre_routing_finish_bridge(struct sk_buff *skb)
 {
        if (skb->pkt_type == PACKET_OTHERHOST) {
@@ -248,32 +303,34 @@ static int br_nf_pre_routing_finish_bridge(struct sk_buff *skb)
        skb->nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
 
        skb->dev = bridge_parent(skb->dev);
-       if (!skb->dev)
-               kfree_skb(skb);
-       else {
-               if (skb->protocol == htons(ETH_P_8021Q)) {
-                       skb_pull(skb, VLAN_HLEN);
-                       skb->nh.raw += VLAN_HLEN;
-               }
-               skb->dst->output(skb);
+       if (skb->dev) {
+               struct dst_entry *dst = skb->dst;
+
+               nf_bridge_pull_encap_header(skb);
+
+               if (dst->hh)
+                       return neigh_hh_output(dst->hh, skb);
+               else if (dst->neighbour)
+                       return dst->neighbour->output(skb);
        }
+       kfree_skb(skb);
        return 0;
 }
 
 static int br_nf_pre_routing_finish(struct sk_buff *skb)
 {
        struct net_device *dev = skb->dev;
-       struct iphdr *iph = skb->nh.iph;
+       struct iphdr *iph = ip_hdr(skb);
        struct nf_bridge_info *nf_bridge = skb->nf_bridge;
+       int err;
 
        if (nf_bridge->mask & BRNF_PKT_TYPE) {
                skb->pkt_type = PACKET_OTHERHOST;
                nf_bridge->mask ^= BRNF_PKT_TYPE;
        }
        nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
-
        if (dnat_took_place(skb)) {
-               if (ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev)) {
+               if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
                        struct rtable *rt;
                        struct flowi fl = {
                                .nl_u = {
@@ -284,19 +341,33 @@ static int br_nf_pre_routing_finish(struct sk_buff *skb)
                                },
                                .proto = 0,
                        };
+                       struct in_device *in_dev = in_dev_get(dev);
+
+                       /* If err equals -EHOSTUNREACH the error is due to a
+                        * martian destination or due to the fact that
+                        * forwarding is disabled. For most martian packets,
+                        * ip_route_output_key() will fail. It won't fail for 2 types of
+                        * martian destinations: loopback destinations and destination
+                        * 0.0.0.0. In both cases the packet will be dropped because the
+                        * destination is the loopback device and not the bridge. */
+                       if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
+                               goto free_skb;
 
                        if (!ip_route_output_key(&rt, &fl)) {
                                /* - Bridged-and-DNAT'ed traffic doesn't
-                                *   require ip_forwarding.
-                                * - Deal with redirected traffic. */
-                               if (((struct dst_entry *)rt)->dev == dev ||
-                                   rt->rt_type == RTN_LOCAL) {
+                                *   require ip_forwarding. */
+                               if (((struct dst_entry *)rt)->dev == dev) {
                                        skb->dst = (struct dst_entry *)rt;
                                        goto bridged_dnat;
                                }
+                               /* we are sure that forwarding is disabled, so printing
+                                * this message is no problem. Note that the packet could
+                                * still have a martian destination address, in which case
+                                * the packet could be dropped even if forwarding were enabled */
                                __br_dnat_complain();
                                dst_release((struct dst_entry *)rt);
                        }
+free_skb:
                        kfree_skb(skb);
                        return 0;
                } else {
@@ -306,11 +377,7 @@ bridged_dnat:
                                 * bridged frame */
                                nf_bridge->mask |= BRNF_BRIDGED_DNAT;
                                skb->dev = nf_bridge->physindev;
-                               if (skb->protocol ==
-                                   htons(ETH_P_8021Q)) {
-                                       skb_push(skb, VLAN_HLEN);
-                                       skb->nh.raw -= VLAN_HLEN;
-                               }
+                               nf_bridge_push_encap_header(skb);
                                NF_HOOK_THRESH(PF_BRIDGE, NF_BR_PRE_ROUTING,
                                               skb, skb->dev, NULL,
                                               br_nf_pre_routing_finish_bridge,
@@ -326,10 +393,7 @@ bridged_dnat:
        }
 
        skb->dev = nf_bridge->physindev;
-       if (skb->protocol == htons(ETH_P_8021Q)) {
-               skb_push(skb, VLAN_HLEN);
-               skb->nh.raw -= VLAN_HLEN;
-       }
+       nf_bridge_push_encap_header(skb);
        NF_HOOK_THRESH(PF_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
                       br_handle_frame_finish, 1);
 
@@ -356,9 +420,10 @@ static struct net_device *setup_pre_routing(struct sk_buff *skb)
 /* We only check the length. A bridge shouldn't do any hop-by-hop stuff anyway */
 static int check_hbh_len(struct sk_buff *skb)
 {
-       unsigned char *raw = (u8 *) (skb->nh.ipv6h + 1);
+       unsigned char *raw = (u8 *)(ipv6_hdr(skb) + 1);
        u32 pkt_len;
-       int off = raw - skb->nh.raw;
+       const unsigned char *nh = skb_network_header(skb);
+       int off = raw - nh;
        int len = (raw[1] + 1) << 3;
 
        if ((raw + len) - skb->data > skb_headlen(skb))
@@ -368,9 +433,9 @@ static int check_hbh_len(struct sk_buff *skb)
        len -= 2;
 
        while (len > 0) {
-               int optlen = skb->nh.raw[off + 1] + 2;
+               int optlen = nh[off + 1] + 2;
 
-               switch (skb->nh.raw[off]) {
+               switch (nh[off]) {
                case IPV6_TLV_PAD0:
                        optlen = 1;
                        break;
@@ -379,17 +444,18 @@ static int check_hbh_len(struct sk_buff *skb)
                        break;
 
                case IPV6_TLV_JUMBO:
-                       if (skb->nh.raw[off + 1] != 4 || (off & 3) != 2)
+                       if (nh[off + 1] != 4 || (off & 3) != 2)
                                goto bad;
-                       pkt_len = ntohl(*(u32 *) (skb->nh.raw + off + 2));
+                       pkt_len = ntohl(*(__be32 *) (nh + off + 2));
                        if (pkt_len <= IPV6_MAXPLEN ||
-                           skb->nh.ipv6h->payload_len)
+                           ipv6_hdr(skb)->payload_len)
                                goto bad;
                        if (pkt_len > skb->len - sizeof(struct ipv6hdr))
                                goto bad;
                        if (pskb_trim_rcsum(skb,
                                            pkt_len + sizeof(struct ipv6hdr)))
                                goto bad;
+                       nh = skb_network_header(skb);
                        break;
                default:
                        if (optlen > len)
@@ -423,7 +489,7 @@ static unsigned int br_nf_pre_routing_ipv6(unsigned int hook,
        if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
                goto inhdr_error;
 
-       hdr = skb->nh.ipv6h;
+       hdr = ipv6_hdr(skb);
 
        if (hdr->version != 6)
                goto inhdr_error;
@@ -445,7 +511,7 @@ static unsigned int br_nf_pre_routing_ipv6(unsigned int hook,
        if (!setup_pre_routing(skb))
                return NF_DROP;
 
-       NF_HOOK(PF_INET6, NF_IP6_PRE_ROUTING, skb, skb->dev, NULL,
+       NF_HOOK(PF_INET6, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
                br_nf_pre_routing_finish_ipv6);
 
        return NF_STOLEN;
@@ -460,27 +526,24 @@ inhdr_error:
  * receiving device) to make netfilter happy, the REDIRECT
  * target in particular.  Save the original destination IP
  * address to be able to detect DNAT afterwards. */
-static unsigned int br_nf_pre_routing(unsigned int hook, struct sk_buff **pskb,
+static unsigned int br_nf_pre_routing(unsigned int hook, struct sk_buff *skb,
                                      const struct net_device *in,
                                      const struct net_device *out,
                                      int (*okfn)(struct sk_buff *))
 {
        struct iphdr *iph;
-       __u32 len;
-       struct sk_buff *skb = *pskb;
+       __u32 len = nf_bridge_encap_header_len(skb);
 
-       if (skb->protocol == htons(ETH_P_IPV6) || IS_VLAN_IPV6(skb)) {
+       if (unlikely(!pskb_may_pull(skb, len)))
+               goto out;
+
+       if (skb->protocol == htons(ETH_P_IPV6) || IS_VLAN_IPV6(skb) ||
+           IS_PPPOE_IPV6(skb)) {
 #ifdef CONFIG_SYSCTL
                if (!brnf_call_ip6tables)
                        return NF_ACCEPT;
 #endif
-               if ((skb = skb_share_check(*pskb, GFP_ATOMIC)) == NULL)
-                       goto out;
-
-               if (skb->protocol == htons(ETH_P_8021Q)) {
-                       skb_pull_rcsum(skb, VLAN_HLEN);
-                       skb->nh.raw += VLAN_HLEN;
-               }
+               nf_bridge_pull_encap_header_rcsum(skb);
                return br_nf_pre_routing_ipv6(hook, skb, in, out, okfn);
        }
 #ifdef CONFIG_SYSCTL
@@ -488,28 +551,23 @@ static unsigned int br_nf_pre_routing(unsigned int hook, struct sk_buff **pskb,
                return NF_ACCEPT;
 #endif
 
-       if (skb->protocol != htons(ETH_P_IP) && !IS_VLAN_IP(skb))
+       if (skb->protocol != htons(ETH_P_IP) && !IS_VLAN_IP(skb) &&
+           !IS_PPPOE_IP(skb))
                return NF_ACCEPT;
 
-       if ((skb = skb_share_check(*pskb, GFP_ATOMIC)) == NULL)
-               goto out;
-
-       if (skb->protocol == htons(ETH_P_8021Q)) {
-               skb_pull_rcsum(skb, VLAN_HLEN);
-               skb->nh.raw += VLAN_HLEN;
-       }
+       nf_bridge_pull_encap_header_rcsum(skb);
 
        if (!pskb_may_pull(skb, sizeof(struct iphdr)))
                goto inhdr_error;
 
-       iph = skb->nh.iph;
+       iph = ip_hdr(skb);
        if (iph->ihl < 5 || iph->version != 4)
                goto inhdr_error;
 
        if (!pskb_may_pull(skb, 4 * iph->ihl))
                goto inhdr_error;
 
-       iph = skb->nh.iph;
+       iph = ip_hdr(skb);
        if (ip_fast_csum((__u8 *) iph, iph->ihl) != 0)
                goto inhdr_error;
 
@@ -526,7 +584,7 @@ static unsigned int br_nf_pre_routing(unsigned int hook, struct sk_buff **pskb,
                return NF_DROP;
        store_orig_dstaddr(skb);
 
-       NF_HOOK(PF_INET, NF_IP_PRE_ROUTING, skb, skb->dev, NULL,
+       NF_HOOK(PF_INET, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
                br_nf_pre_routing_finish);
 
        return NF_STOLEN;
@@ -545,13 +603,11 @@ out:
  * took place when the packet entered the bridge), but we
  * register an IPv4 PRE_ROUTING 'sabotage' hook that will
  * prevent this from happening. */
-static unsigned int br_nf_local_in(unsigned int hook, struct sk_buff **pskb,
+static unsigned int br_nf_local_in(unsigned int hook, struct sk_buff *skb,
                                   const struct net_device *in,
                                   const struct net_device *out,
                                   int (*okfn)(struct sk_buff *))
 {
-       struct sk_buff *skb = *pskb;
-
        if (skb->dst == (struct dst_entry *)&__fake_rtable) {
                dst_release(skb->dst);
                skb->dst = NULL;
@@ -575,10 +631,7 @@ static int br_nf_forward_finish(struct sk_buff *skb)
        } else {
                in = *((struct net_device **)(skb->cb));
        }
-       if (skb->protocol == htons(ETH_P_8021Q)) {
-               skb_push(skb, VLAN_HLEN);
-               skb->nh.raw -= VLAN_HLEN;
-       }
+       nf_bridge_push_encap_header(skb);
        NF_HOOK_THRESH(PF_BRIDGE, NF_BR_FORWARD, skb, in,
                       skb->dev, br_forward_finish, 1);
        return 0;
@@ -589,12 +642,11 @@ static int br_nf_forward_finish(struct sk_buff *skb)
  * but we are still able to filter on the 'real' indev/outdev
  * because of the physdev module. For ARP, indev and outdev are the
  * bridge ports. */
-static unsigned int br_nf_forward_ip(unsigned int hook, struct sk_buff **pskb,
+static unsigned int br_nf_forward_ip(unsigned int hook, struct sk_buff *skb,
                                     const struct net_device *in,
                                     const struct net_device *out,
                                     int (*okfn)(struct sk_buff *))
 {
-       struct sk_buff *skb = *pskb;
        struct nf_bridge_info *nf_bridge;
        struct net_device *parent;
        int pf;
@@ -602,19 +654,22 @@ static unsigned int br_nf_forward_ip(unsigned int hook, struct sk_buff **pskb,
        if (!skb->nf_bridge)
                return NF_ACCEPT;
 
+       /* Need exclusive nf_bridge_info since we might have multiple
+        * different physoutdevs. */
+       if (!nf_bridge_unshare(skb))
+               return NF_DROP;
+
        parent = bridge_parent(out);
        if (!parent)
                return NF_DROP;
 
-       if (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb))
+       if (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb) ||
+           IS_PPPOE_IP(skb))
                pf = PF_INET;
        else
                pf = PF_INET6;
 
-       if (skb->protocol == htons(ETH_P_8021Q)) {
-               skb_pull(*pskb, VLAN_HLEN);
-               (*pskb)->nh.raw += VLAN_HLEN;
-       }
+       nf_bridge_pull_encap_header(skb);
 
        nf_bridge = skb->nf_bridge;
        if (skb->pkt_type == PACKET_OTHERHOST) {
@@ -626,18 +681,17 @@ static unsigned int br_nf_forward_ip(unsigned int hook, struct sk_buff **pskb,
        nf_bridge->mask |= BRNF_BRIDGED;
        nf_bridge->physoutdev = skb->dev;
 
-       NF_HOOK(pf, NF_IP_FORWARD, skb, bridge_parent(in), parent,
+       NF_HOOK(pf, NF_INET_FORWARD, skb, bridge_parent(in), parent,
                br_nf_forward_finish);
 
        return NF_STOLEN;
 }
 
-static unsigned int br_nf_forward_arp(unsigned int hook, struct sk_buff **pskb,
+static unsigned int br_nf_forward_arp(unsigned int hook, struct sk_buff *skb,
                                      const struct net_device *in,
                                      const struct net_device *out,
                                      int (*okfn)(struct sk_buff *))
 {
-       struct sk_buff *skb = *pskb;
        struct net_device **d = (struct net_device **)(skb->cb);
 
 #ifdef CONFIG_SYSCTL
@@ -648,15 +702,12 @@ static unsigned int br_nf_forward_arp(unsigned int hook, struct sk_buff **pskb,
        if (skb->protocol != htons(ETH_P_ARP)) {
                if (!IS_VLAN_ARP(skb))
                        return NF_ACCEPT;
-               skb_pull(*pskb, VLAN_HLEN);
-               (*pskb)->nh.raw += VLAN_HLEN;
+               nf_bridge_pull_encap_header(skb);
        }
 
-       if (skb->nh.arph->ar_pln != 4) {
-               if (IS_VLAN_ARP(skb)) {
-                       skb_push(*pskb, VLAN_HLEN);
-                       (*pskb)->nh.raw -= VLAN_HLEN;
-               }
+       if (arp_hdr(skb)->ar_pln != 4) {
+               if (IS_VLAN_ARP(skb))
+                       nf_bridge_push_encap_header(skb);
                return NF_ACCEPT;
        }
        *d = (struct net_device *)in;
@@ -666,110 +717,51 @@ static unsigned int br_nf_forward_arp(unsigned int hook, struct sk_buff **pskb,
        return NF_STOLEN;
 }
 
-/* PF_BRIDGE/LOCAL_OUT ***********************************************/
-static int br_nf_local_out_finish(struct sk_buff *skb)
-{
-       if (skb->protocol == htons(ETH_P_8021Q)) {
-               skb_push(skb, VLAN_HLEN);
-               skb->nh.raw -= VLAN_HLEN;
-       }
-
-       NF_HOOK_THRESH(PF_BRIDGE, NF_BR_LOCAL_OUT, skb, NULL, skb->dev,
-                      br_forward_finish, NF_BR_PRI_FIRST + 1);
-
-       return 0;
-}
-
-/* This function sees both locally originated IP packets and forwarded
+/* PF_BRIDGE/LOCAL_OUT ***********************************************
+ *
+ * This function sees both locally originated IP packets and forwarded
  * IP packets (in both cases the destination device is a bridge
  * device). It also sees bridged-and-DNAT'ed packets.
- * To be able to filter on the physical bridge devices (with the physdev
- * module), we steal packets destined to a bridge device away from the
- * PF_INET/FORWARD and PF_INET/OUTPUT hook functions, and give them back later,
- * when we have determined the real output device. This is done in here.
  *
  * If (nf_bridge->mask & BRNF_BRIDGED_DNAT) then the packet is bridged
  * and we fake the PF_BRIDGE/FORWARD hook. The function br_nf_forward()
  * will then fake the PF_INET/FORWARD hook. br_nf_local_out() has priority
  * NF_BR_PRI_FIRST, so no relevant PF_BRIDGE/INPUT functions have been nor
  * will be executed.
- * Otherwise, if nf_bridge->physindev is NULL, the bridge-nf code never touched
- * this packet before, and so the packet was locally originated. We fake
- * the PF_INET/LOCAL_OUT hook.
- * Finally, if nf_bridge->physindev isn't NULL, then the packet was IP routed,
- * so we fake the PF_INET/FORWARD hook. ip_sabotage_out() makes sure
- * even routed packets that didn't arrive on a bridge interface have their
- * nf_bridge->physindev set. */
-static unsigned int br_nf_local_out(unsigned int hook, struct sk_buff **pskb,
+ */
+static unsigned int br_nf_local_out(unsigned int hook, struct sk_buff *skb,
                                    const struct net_device *in,
                                    const struct net_device *out,
                                    int (*okfn)(struct sk_buff *))
 {
-       struct net_device *realindev, *realoutdev;
-       struct sk_buff *skb = *pskb;
+       struct net_device *realindev;
        struct nf_bridge_info *nf_bridge;
-       int pf;
 
        if (!skb->nf_bridge)
                return NF_ACCEPT;
 
-       if (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb))
-               pf = PF_INET;
-       else
-               pf = PF_INET6;
+       /* Need exclusive nf_bridge_info since we might have multiple
+        * different physoutdevs. */
+       if (!nf_bridge_unshare(skb))
+               return NF_DROP;
 
        nf_bridge = skb->nf_bridge;
-       nf_bridge->physoutdev = skb->dev;
-       realindev = nf_bridge->physindev;
+       if (!(nf_bridge->mask & BRNF_BRIDGED_DNAT))
+               return NF_ACCEPT;
 
        /* Bridged, take PF_BRIDGE/FORWARD.
         * (see big note in front of br_nf_pre_routing_finish) */
-       if (nf_bridge->mask & BRNF_BRIDGED_DNAT) {
-               if (nf_bridge->mask & BRNF_PKT_TYPE) {
-                       skb->pkt_type = PACKET_OTHERHOST;
-                       nf_bridge->mask ^= BRNF_PKT_TYPE;
-               }
-               if (skb->protocol == htons(ETH_P_8021Q)) {
-                       skb_push(skb, VLAN_HLEN);
-                       skb->nh.raw -= VLAN_HLEN;
-               }
-
-               NF_HOOK(PF_BRIDGE, NF_BR_FORWARD, skb, realindev,
-                       skb->dev, br_forward_finish);
-               goto out;
-       }
-       realoutdev = bridge_parent(skb->dev);
-       if (!realoutdev)
-               return NF_DROP;
-
-#if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
-       /* iptables should match -o br0.x */
-       if (nf_bridge->netoutdev)
-               realoutdev = nf_bridge->netoutdev;
-#endif
-       if (skb->protocol == htons(ETH_P_8021Q)) {
-               skb_pull(skb, VLAN_HLEN);
-               (*pskb)->nh.raw += VLAN_HLEN;
-       }
-       /* IP forwarded traffic has a physindev, locally
-        * generated traffic hasn't. */
-       if (realindev != NULL) {
-               if (!(nf_bridge->mask & BRNF_DONT_TAKE_PARENT)) {
-                       struct net_device *parent = bridge_parent(realindev);
-                       if (parent)
-                               realindev = parent;
-               }
+       nf_bridge->physoutdev = skb->dev;
+       realindev = nf_bridge->physindev;
 
-               NF_HOOK_THRESH(pf, NF_IP_FORWARD, skb, realindev,
-                              realoutdev, br_nf_local_out_finish,
-                              NF_IP_PRI_BRIDGE_SABOTAGE_FORWARD + 1);
-       } else {
-               NF_HOOK_THRESH(pf, NF_IP_LOCAL_OUT, skb, realindev,
-                              realoutdev, br_nf_local_out_finish,
-                              NF_IP_PRI_BRIDGE_SABOTAGE_LOCAL_OUT + 1);
+       if (nf_bridge->mask & BRNF_PKT_TYPE) {
+               skb->pkt_type = PACKET_OTHERHOST;
+               nf_bridge->mask ^= BRNF_PKT_TYPE;
        }
+       nf_bridge_push_encap_header(skb);
 
-out:
+       NF_HOOK(PF_BRIDGE, NF_BR_FORWARD, skb, realindev, skb->dev,
+               br_forward_finish);
        return NF_STOLEN;
 }
 
@@ -784,20 +776,20 @@ static int br_nf_dev_queue_xmit(struct sk_buff *skb)
 }
 
 /* PF_BRIDGE/POST_ROUTING ********************************************/
-static unsigned int br_nf_post_routing(unsigned int hook, struct sk_buff **pskb,
+static unsigned int br_nf_post_routing(unsigned int hook, struct sk_buff *skb,
                                       const struct net_device *in,
                                       const struct net_device *out,
                                       int (*okfn)(struct sk_buff *))
 {
-       struct sk_buff *skb = *pskb;
-       struct nf_bridge_info *nf_bridge = (*pskb)->nf_bridge;
+       struct nf_bridge_info *nf_bridge = skb->nf_bridge;
        struct net_device *realoutdev = bridge_parent(skb->dev);
        int pf;
 
 #ifdef CONFIG_NETFILTER_DEBUG
        /* Be very paranoid. This probably won't happen anymore, but let's
         * keep the check just to be sure... */
-       if (skb->mac.raw < skb->head || skb->mac.raw + ETH_HLEN > skb->data) {
+       if (skb_mac_header(skb) < skb->head ||
+           skb_mac_header(skb) + ETH_HLEN > skb->data) {
                printk(KERN_CRIT "br_netfilter: Argh!! br_nf_post_routing: "
                       "bad mac.raw pointer.\n");
                goto print_error;
@@ -807,10 +799,14 @@ static unsigned int br_nf_post_routing(unsigned int hook, struct sk_buff **pskb,
        if (!nf_bridge)
                return NF_ACCEPT;
 
+       if (!(nf_bridge->mask & (BRNF_BRIDGED | BRNF_BRIDGED_DNAT)))
+               return NF_ACCEPT;
+
        if (!realoutdev)
                return NF_DROP;
 
-       if (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb))
+       if (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb) ||
+           IS_PPPOE_IP(skb))
                pf = PF_INET;
        else
                pf = PF_INET6;
@@ -829,18 +825,14 @@ static unsigned int br_nf_post_routing(unsigned int hook, struct sk_buff **pskb,
                nf_bridge->mask |= BRNF_PKT_TYPE;
        }
 
-       if (skb->protocol == htons(ETH_P_8021Q)) {
-               skb_pull(skb, VLAN_HLEN);
-               skb->nh.raw += VLAN_HLEN;
-       }
-
+       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_IP_POST_ROUTING, skb, NULL, realoutdev,
+       NF_HOOK(pf, NF_INET_POST_ROUTING, skb, NULL, realoutdev,
                br_nf_dev_queue_xmit);
 
        return NF_STOLEN;
@@ -852,7 +844,7 @@ print_error:
                if (realoutdev)
                        printk("[%s]", realoutdev->name);
        }
-       printk(" head:%p, raw:%p, data:%p\n", skb->head, skb->mac.raw,
+       printk(" head:%p, raw:%p, data:%p\n", skb->head, skb_mac_header(skb),
               skb->data);
        dump_stack();
        return NF_ACCEPT;
@@ -862,76 +854,13 @@ print_error:
 /* IP/SABOTAGE *****************************************************/
 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
  * for the second time. */
-static unsigned int ip_sabotage_in(unsigned int hook, struct sk_buff **pskb,
+static unsigned int ip_sabotage_in(unsigned int hook, struct sk_buff *skb,
                                   const struct net_device *in,
                                   const struct net_device *out,
                                   int (*okfn)(struct sk_buff *))
 {
-       if ((*pskb)->nf_bridge &&
-           !((*pskb)->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)) {
-               return NF_STOP;
-       }
-
-       return NF_ACCEPT;
-}
-
-/* Postpone execution of PF_INET(6)/FORWARD, PF_INET(6)/LOCAL_OUT
- * and PF_INET(6)/POST_ROUTING until we have done the forwarding
- * decision in the bridge code and have determined nf_bridge->physoutdev. */
-static unsigned int ip_sabotage_out(unsigned int hook, struct sk_buff **pskb,
-                                   const struct net_device *in,
-                                   const struct net_device *out,
-                                   int (*okfn)(struct sk_buff *))
-{
-       struct sk_buff *skb = *pskb;
-
-       if ((out->hard_start_xmit == br_dev_xmit &&
-            okfn != br_nf_forward_finish &&
-            okfn != br_nf_local_out_finish && okfn != br_nf_dev_queue_xmit)
-#if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
-           || ((out->priv_flags & IFF_802_1Q_VLAN) &&
-               VLAN_DEV_INFO(out)->real_dev->hard_start_xmit == br_dev_xmit)
-#endif
-           ) {
-               struct nf_bridge_info *nf_bridge;
-
-               if (!skb->nf_bridge) {
-#ifdef CONFIG_SYSCTL
-                       /* This code is executed while in the IP(v6) stack,
-                          the version should be 4 or 6. We can't use
-                          skb->protocol because that isn't set on
-                          PF_INET(6)/LOCAL_OUT. */
-                       struct iphdr *ip = skb->nh.iph;
-
-                       if (ip->version == 4 && !brnf_call_iptables)
-                               return NF_ACCEPT;
-                       else if (ip->version == 6 && !brnf_call_ip6tables)
-                               return NF_ACCEPT;
-                       else if (!brnf_deferred_hooks)
-                               return NF_ACCEPT;
-#endif
-                       if (hook == NF_IP_POST_ROUTING)
-                               return NF_ACCEPT;
-                       if (!nf_bridge_alloc(skb))
-                               return NF_DROP;
-               }
-
-               nf_bridge = skb->nf_bridge;
-
-               /* This frame will arrive on PF_BRIDGE/LOCAL_OUT and we
-                * will need the indev then. For a brouter, the real indev
-                * can be a bridge port, so we make sure br_nf_local_out()
-                * doesn't use the bridge parent of the indev by using
-                * the BRNF_DONT_TAKE_PARENT mask. */
-               if (hook == NF_IP_FORWARD && nf_bridge->physindev == NULL) {
-                       nf_bridge->mask |= BRNF_DONT_TAKE_PARENT;
-                       nf_bridge->physindev = (struct net_device *)in;
-               }
-#if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
-               /* the iptables outdev is br0.x, not br0 */
-               if (out->priv_flags & IFF_802_1Q_VLAN)
-                       nf_bridge->netoutdev = (struct net_device *)out;
-#endif
+       if (skb->nf_bridge &&
+           !(skb->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)) {
                return NF_STOP;
        }
 
@@ -943,10 +872,10 @@ static unsigned int ip_sabotage_out(unsigned int hook, struct sk_buff **pskb,
  * For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
  * ip_refrag() can return NF_STOLEN. */
 static struct nf_hook_ops br_nf_ops[] = {
-       { .hook = br_nf_pre_routing, 
-         .owner = THIS_MODULE, 
-         .pf = PF_BRIDGE, 
-         .hooknum = NF_BR_PRE_ROUTING, 
+       { .hook = br_nf_pre_routing,
+         .owner = THIS_MODULE,
+         .pf = PF_BRIDGE,
+         .hooknum = NF_BR_PRE_ROUTING,
          .priority = NF_BR_PRI_BRNF, },
        { .hook = br_nf_local_in,
          .owner = THIS_MODULE,
@@ -976,42 +905,12 @@ static struct nf_hook_ops br_nf_ops[] = {
        { .hook = ip_sabotage_in,
          .owner = THIS_MODULE,
          .pf = PF_INET,
-         .hooknum = NF_IP_PRE_ROUTING,
+         .hooknum = NF_INET_PRE_ROUTING,
          .priority = NF_IP_PRI_FIRST, },
        { .hook = ip_sabotage_in,
          .owner = THIS_MODULE,
          .pf = PF_INET6,
-         .hooknum = NF_IP6_PRE_ROUTING,
-         .priority = NF_IP6_PRI_FIRST, },
-       { .hook = ip_sabotage_out,
-         .owner = THIS_MODULE,
-         .pf = PF_INET,
-         .hooknum = NF_IP_FORWARD,
-         .priority = NF_IP_PRI_BRIDGE_SABOTAGE_FORWARD, },
-       { .hook = ip_sabotage_out,
-         .owner = THIS_MODULE,
-         .pf = PF_INET6,
-         .hooknum = NF_IP6_FORWARD,
-         .priority = NF_IP6_PRI_BRIDGE_SABOTAGE_FORWARD, },
-       { .hook = ip_sabotage_out,
-         .owner = THIS_MODULE,
-         .pf = PF_INET,
-         .hooknum = NF_IP_LOCAL_OUT,
-         .priority = NF_IP_PRI_BRIDGE_SABOTAGE_LOCAL_OUT, },
-       { .hook = ip_sabotage_out,
-         .owner = THIS_MODULE,
-         .pf = PF_INET6,
-         .hooknum = NF_IP6_LOCAL_OUT,
-         .priority = NF_IP6_PRI_BRIDGE_SABOTAGE_LOCAL_OUT, },
-       { .hook = ip_sabotage_out,
-         .owner = THIS_MODULE,
-         .pf = PF_INET,
-         .hooknum = NF_IP_POST_ROUTING,
-         .priority = NF_IP_PRI_FIRST, },
-       { .hook = ip_sabotage_out,
-         .owner = THIS_MODULE,
-         .pf = PF_INET6,
-         .hooknum = NF_IP6_POST_ROUTING,
+         .hooknum = NF_INET_PRE_ROUTING,
          .priority = NF_IP6_PRI_FIRST, },
 };
 
@@ -1031,7 +930,6 @@ int brnf_sysctl_call_tables(ctl_table * ctl, int write, struct file *filp,
 
 static ctl_table brnf_table[] = {
        {
-               .ctl_name       = NET_BRIDGE_NF_CALL_ARPTABLES,
                .procname       = "bridge-nf-call-arptables",
                .data           = &brnf_call_arptables,
                .maxlen         = sizeof(int),
@@ -1039,7 +937,6 @@ static ctl_table brnf_table[] = {
                .proc_handler   = &brnf_sysctl_call_tables,
        },
        {
-               .ctl_name       = NET_BRIDGE_NF_CALL_IPTABLES,
                .procname       = "bridge-nf-call-iptables",
                .data           = &brnf_call_iptables,
                .maxlen         = sizeof(int),
@@ -1047,7 +944,6 @@ static ctl_table brnf_table[] = {
                .proc_handler   = &brnf_sysctl_call_tables,
        },
        {
-               .ctl_name       = NET_BRIDGE_NF_CALL_IP6TABLES,
                .procname       = "bridge-nf-call-ip6tables",
                .data           = &brnf_call_ip6tables,
                .maxlen         = sizeof(int),
@@ -1055,13 +951,19 @@ static ctl_table brnf_table[] = {
                .proc_handler   = &brnf_sysctl_call_tables,
        },
        {
-               .ctl_name       = NET_BRIDGE_NF_FILTER_VLAN_TAGGED,
                .procname       = "bridge-nf-filter-vlan-tagged",
                .data           = &brnf_filter_vlan_tagged,
                .maxlen         = sizeof(int),
                .mode           = 0644,
                .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,
+       },
        { .ctl_name = 0 }
 };
 
@@ -1086,44 +988,29 @@ static ctl_table brnf_net_table[] = {
 };
 #endif
 
-int br_netfilter_init(void)
+int __init br_netfilter_init(void)
 {
-       int i;
-
-       for (i = 0; i < ARRAY_SIZE(br_nf_ops); i++) {
-               int ret;
-
-               if ((ret = nf_register_hook(&br_nf_ops[i])) >= 0)
-                       continue;
-
-               while (i--)
-                       nf_unregister_hook(&br_nf_ops[i]);
+       int ret;
 
+       ret = nf_register_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
+       if (ret < 0)
                return ret;
-       }
-
 #ifdef CONFIG_SYSCTL
-       brnf_sysctl_header = register_sysctl_table(brnf_net_table, 0);
+       brnf_sysctl_header = register_sysctl_table(brnf_net_table);
        if (brnf_sysctl_header == NULL) {
                printk(KERN_WARNING
                       "br_netfilter: can't register to sysctl.\n");
-               for (i = 0; i < ARRAY_SIZE(br_nf_ops); i++)
-                       nf_unregister_hook(&br_nf_ops[i]);
-               return -EFAULT;
+               nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
+               return -ENOMEM;
        }
 #endif
-
        printk(KERN_NOTICE "Bridge firewalling registered\n");
-
        return 0;
 }
 
 void br_netfilter_fini(void)
 {
-       int i;
-
-       for (i = ARRAY_SIZE(br_nf_ops) - 1; i >= 0; i--)
-               nf_unregister_hook(&br_nf_ops[i]);
+       nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
 #ifdef CONFIG_SYSCTL
        unregister_sysctl_table(brnf_sysctl_header);
 #endif