Merge branch 'nf' of git://1984.lsi.us.es/net
[linux-2.6.git] / net / ipv4 / netfilter.c
index 599d448..4f47e06 100644 (file)
@@ -4,71 +4,68 @@
 #include <linux/netfilter_ipv4.h>
 #include <linux/ip.h>
 #include <linux/skbuff.h>
+#include <linux/gfp.h>
+#include <linux/export.h>
 #include <net/route.h>
 #include <net/xfrm.h>
 #include <net/ip.h>
+#include <net/netfilter/nf_queue.h>
 
 /* route_me_harder function, used by iptable_nat, iptable_mangle + ip_queue */
 int ip_route_me_harder(struct sk_buff *skb, unsigned addr_type)
 {
+       struct net *net = dev_net(skb_dst(skb)->dev);
        const struct iphdr *iph = ip_hdr(skb);
        struct rtable *rt;
-       struct flowi fl = {};
-       struct dst_entry *odst;
+       struct flowi4 fl4 = {};
+       __be32 saddr = iph->saddr;
+       __u8 flags = skb->sk ? inet_sk_flowi_flags(skb->sk) : 0;
        unsigned int hh_len;
-       unsigned int type;
 
-       type = inet_addr_type(iph->saddr);
        if (addr_type == RTN_UNSPEC)
-               addr_type = type;
+               addr_type = inet_addr_type(net, saddr);
+       if (addr_type == RTN_LOCAL || addr_type == RTN_UNICAST)
+               flags |= FLOWI_FLAG_ANYSRC;
+       else
+               saddr = 0;
 
        /* some non-standard hacks like ipt_REJECT.c:send_reset() can cause
         * packets with foreign saddr to appear on the NF_INET_LOCAL_OUT hook.
         */
-       if (addr_type == RTN_LOCAL) {
-               fl.nl_u.ip4_u.daddr = iph->daddr;
-               if (type == RTN_LOCAL)
-                       fl.nl_u.ip4_u.saddr = iph->saddr;
-               fl.nl_u.ip4_u.tos = RT_TOS(iph->tos);
-               fl.oif = skb->sk ? skb->sk->sk_bound_dev_if : 0;
-               fl.mark = skb->mark;
-               if (ip_route_output_key(&rt, &fl) != 0)
-                       return -1;
-
-               /* Drop old route. */
-               dst_release(skb->dst);
-               skb->dst = &rt->u.dst;
-       } else {
-               /* non-local src, find valid iif to satisfy
-                * rp-filter when calling ip_route_input. */
-               fl.nl_u.ip4_u.daddr = iph->saddr;
-               if (ip_route_output_key(&rt, &fl) != 0)
-                       return -1;
+       fl4.daddr = iph->daddr;
+       fl4.saddr = saddr;
+       fl4.flowi4_tos = RT_TOS(iph->tos);
+       fl4.flowi4_oif = skb->sk ? skb->sk->sk_bound_dev_if : 0;
+       fl4.flowi4_mark = skb->mark;
+       fl4.flowi4_flags = flags;
+       rt = ip_route_output_key(net, &fl4);
+       if (IS_ERR(rt))
+               return -1;
 
-               odst = skb->dst;
-               if (ip_route_input(skb, iph->daddr, iph->saddr,
-                                  RT_TOS(iph->tos), rt->u.dst.dev) != 0) {
-                       dst_release(&rt->u.dst);
-                       return -1;
-               }
-               dst_release(&rt->u.dst);
-               dst_release(odst);
-       }
+       /* Drop old route. */
+       skb_dst_drop(skb);
+       skb_dst_set(skb, &rt->dst);
 
-       if (skb->dst->error)
+       if (skb_dst(skb)->error)
                return -1;
 
 #ifdef CONFIG_XFRM
        if (!(IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED) &&
-           xfrm_decode_session(skb, &fl, AF_INET) == 0)
-               if (xfrm_lookup(&skb->dst, &fl, skb->sk, 0))
+           xfrm_decode_session(skb, flowi4_to_flowi(&fl4), AF_INET) == 0) {
+               struct dst_entry *dst = skb_dst(skb);
+               skb_dst_set(skb, NULL);
+               dst = xfrm_lookup(net, dst, flowi4_to_flowi(&fl4), skb->sk, 0);
+               if (IS_ERR(dst))
                        return -1;
+               skb_dst_set(skb, dst);
+       }
 #endif
 
        /* Change in oif may mean change in hh_len. */
-       hh_len = skb->dst->dev->hard_header_len;
+       hh_len = skb_dst(skb)->dev->hard_header_len;
        if (skb_headroom(skb) < hh_len &&
-           pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
+           pskb_expand_head(skb, HH_DATA_ALIGN(hh_len - skb_headroom(skb)),
+                               0, GFP_ATOMIC))
                return -1;
 
        return 0;
@@ -87,19 +84,20 @@ int ip_xfrm_me_harder(struct sk_buff *skb)
        if (xfrm_decode_session(skb, &fl, AF_INET) < 0)
                return -1;
 
-       dst = skb->dst;
+       dst = skb_dst(skb);
        if (dst->xfrm)
                dst = ((struct xfrm_dst *)dst)->route;
        dst_hold(dst);
 
-       if (xfrm_lookup(&dst, &fl, skb->sk, 0) < 0)
+       dst = xfrm_lookup(dev_net(dst->dev), dst, &fl, skb->sk, 0);
+       if (IS_ERR(dst))
                return -1;
 
-       dst_release(skb->dst);
-       skb->dst = dst;
+       skb_dst_drop(skb);
+       skb_dst_set(skb, dst);
 
        /* Change in oif may mean change in hh_len. */
-       hh_len = skb->dst->dev->hard_header_len;
+       hh_len = skb_dst(skb)->dev->hard_header_len;
        if (skb_headroom(skb) < hh_len &&
            pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
                return -1;
@@ -120,31 +118,36 @@ struct ip_rt_info {
        __be32 daddr;
        __be32 saddr;
        u_int8_t tos;
+       u_int32_t mark;
 };
 
-static void nf_ip_saveroute(const struct sk_buff *skb, struct nf_info *info)
+static void nf_ip_saveroute(const struct sk_buff *skb,
+                           struct nf_queue_entry *entry)
 {
-       struct ip_rt_info *rt_info = nf_info_reroute(info);
+       struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
 
-       if (info->hook == NF_INET_LOCAL_OUT) {
+       if (entry->hook == NF_INET_LOCAL_OUT) {
                const struct iphdr *iph = ip_hdr(skb);
 
                rt_info->tos = iph->tos;
                rt_info->daddr = iph->daddr;
                rt_info->saddr = iph->saddr;
+               rt_info->mark = skb->mark;
        }
 }
 
-static int nf_ip_reroute(struct sk_buff *skb, const struct nf_info *info)
+static int nf_ip_reroute(struct sk_buff *skb,
+                        const struct nf_queue_entry *entry)
 {
-       const struct ip_rt_info *rt_info = nf_info_reroute(info);
+       const struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
 
-       if (info->hook == NF_INET_LOCAL_OUT) {
+       if (entry->hook == NF_INET_LOCAL_OUT) {
                const struct iphdr *iph = ip_hdr(skb);
 
-               if (!(iph->tos == rt_info->tos
-                     && iph->daddr == rt_info->daddr
-                     && iph->saddr == rt_info->saddr))
+               if (!(iph->tos == rt_info->tos &&
+                     skb->mark == rt_info->mark &&
+                     iph->daddr == rt_info->daddr &&
+                     iph->saddr == rt_info->saddr))
                        return ip_route_me_harder(skb, RTN_UNSPEC);
        }
        return 0;
@@ -179,21 +182,47 @@ __sum16 nf_ip_checksum(struct sk_buff *skb, unsigned int hook,
        }
        return csum;
 }
-
 EXPORT_SYMBOL(nf_ip_checksum);
 
-static int nf_ip_route(struct dst_entry **dst, struct flowi *fl)
+static __sum16 nf_ip_checksum_partial(struct sk_buff *skb, unsigned int hook,
+                                     unsigned int dataoff, unsigned int len,
+                                     u_int8_t protocol)
 {
-       return ip_route_output_key((struct rtable **)dst, fl);
+       const struct iphdr *iph = ip_hdr(skb);
+       __sum16 csum = 0;
+
+       switch (skb->ip_summed) {
+       case CHECKSUM_COMPLETE:
+               if (len == skb->len - dataoff)
+                       return nf_ip_checksum(skb, hook, dataoff, protocol);
+               /* fall through */
+       case CHECKSUM_NONE:
+               skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, protocol,
+                                              skb->len - dataoff, 0);
+               skb->ip_summed = CHECKSUM_NONE;
+               return __skb_checksum_complete_head(skb, dataoff + len);
+       }
+       return csum;
 }
 
-static struct nf_afinfo nf_ip_afinfo = {
-       .family         = AF_INET,
-       .checksum       = nf_ip_checksum,
-       .route          = nf_ip_route,
-       .saveroute      = nf_ip_saveroute,
-       .reroute        = nf_ip_reroute,
-       .route_key_size = sizeof(struct ip_rt_info),
+static int nf_ip_route(struct net *net, struct dst_entry **dst,
+                      struct flowi *fl, bool strict __always_unused)
+{
+       struct rtable *rt = ip_route_output_key(net, &fl->u.ip4);
+       if (IS_ERR(rt))
+               return PTR_ERR(rt);
+       *dst = &rt->dst;
+       return 0;
+}
+
+static const struct nf_afinfo nf_ip_afinfo = {
+       .family                 = AF_INET,
+       .checksum               = nf_ip_checksum,
+       .checksum_partial       = nf_ip_checksum_partial,
+       .route                  = nf_ip_route,
+       .saveroute              = nf_ip_saveroute,
+       .reroute                = nf_ip_reroute,
+       .route_key_size         = sizeof(struct ip_rt_info),
 };
 
 static int ipv4_netfilter_init(void)
@@ -208,3 +237,13 @@ static void ipv4_netfilter_fini(void)
 
 module_init(ipv4_netfilter_init);
 module_exit(ipv4_netfilter_fini);
+
+#ifdef CONFIG_SYSCTL
+struct ctl_path nf_net_ipv4_netfilter_sysctl_path[] = {
+       { .procname = "net", },
+       { .procname = "ipv4", },
+       { .procname = "netfilter", },
+       { }
+};
+EXPORT_SYMBOL_GPL(nf_net_ipv4_netfilter_sysctl_path);
+#endif /* CONFIG_SYSCTL */