Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / net / ipv4 / ip_forward.c
diff --git a/kernel/net/ipv4/ip_forward.c b/kernel/net/ipv4/ip_forward.c
new file mode 100644 (file)
index 0000000..3674484
--- /dev/null
@@ -0,0 +1,159 @@
+/*
+ * INET                An implementation of the TCP/IP protocol suite for the LINUX
+ *             operating system.  INET is implemented using the  BSD Socket
+ *             interface as the means of communication with the user level.
+ *
+ *             The IP forwarding functionality.
+ *
+ * Authors:    see ip.c
+ *
+ * Fixes:
+ *             Many            :       Split from ip.c , see ip_input.c for
+ *                                     history.
+ *             Dave Gregorich  :       NULL ip_rt_put fix for multicast
+ *                                     routing.
+ *             Jos Vos         :       Add call_out_firewall before sending,
+ *                                     use output device for accounting.
+ *             Jos Vos         :       Call forward firewall after routing
+ *                                     (always use output device).
+ *             Mike McLagan    :       Routing by source
+ */
+
+#include <linux/types.h>
+#include <linux/mm.h>
+#include <linux/skbuff.h>
+#include <linux/ip.h>
+#include <linux/icmp.h>
+#include <linux/netdevice.h>
+#include <linux/slab.h>
+#include <net/sock.h>
+#include <net/ip.h>
+#include <net/tcp.h>
+#include <net/udp.h>
+#include <net/icmp.h>
+#include <linux/tcp.h>
+#include <linux/udp.h>
+#include <linux/netfilter_ipv4.h>
+#include <net/checksum.h>
+#include <linux/route.h>
+#include <net/route.h>
+#include <net/xfrm.h>
+
+static bool ip_may_fragment(const struct sk_buff *skb)
+{
+       return unlikely((ip_hdr(skb)->frag_off & htons(IP_DF)) == 0) ||
+               skb->ignore_df;
+}
+
+static bool ip_exceeds_mtu(const struct sk_buff *skb, unsigned int mtu)
+{
+       if (skb->len <= mtu)
+               return false;
+
+       if (skb_is_gso(skb) && skb_gso_network_seglen(skb) <= mtu)
+               return false;
+
+       return true;
+}
+
+
+static int ip_forward_finish(struct sock *sk, struct sk_buff *skb)
+{
+       struct ip_options *opt  = &(IPCB(skb)->opt);
+
+       IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTFORWDATAGRAMS);
+       IP_ADD_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTOCTETS, skb->len);
+
+       if (unlikely(opt->optlen))
+               ip_forward_options(skb);
+
+       skb_sender_cpu_clear(skb);
+       return dst_output_sk(sk, skb);
+}
+
+int ip_forward(struct sk_buff *skb)
+{
+       u32 mtu;
+       struct iphdr *iph;      /* Our header */
+       struct rtable *rt;      /* Route we use */
+       struct ip_options *opt  = &(IPCB(skb)->opt);
+
+       /* that should never happen */
+       if (skb->pkt_type != PACKET_HOST)
+               goto drop;
+
+       if (unlikely(skb->sk))
+               goto drop;
+
+       if (skb_warn_if_lro(skb))
+               goto drop;
+
+       if (!xfrm4_policy_check(NULL, XFRM_POLICY_FWD, skb))
+               goto drop;
+
+       if (IPCB(skb)->opt.router_alert && ip_call_ra_chain(skb))
+               return NET_RX_SUCCESS;
+
+       skb_forward_csum(skb);
+
+       /*
+        *      According to the RFC, we must first decrease the TTL field. If
+        *      that reaches zero, we must reply an ICMP control message telling
+        *      that the packet's lifetime expired.
+        */
+       if (ip_hdr(skb)->ttl <= 1)
+               goto too_many_hops;
+
+       if (!xfrm4_route_forward(skb))
+               goto drop;
+
+       rt = skb_rtable(skb);
+
+       if (opt->is_strictroute && rt->rt_uses_gateway)
+               goto sr_failed;
+
+       IPCB(skb)->flags |= IPSKB_FORWARDED;
+       mtu = ip_dst_mtu_maybe_forward(&rt->dst, true);
+       if (!ip_may_fragment(skb) && ip_exceeds_mtu(skb, mtu)) {
+               IP_INC_STATS(dev_net(rt->dst.dev), IPSTATS_MIB_FRAGFAILS);
+               icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
+                         htonl(mtu));
+               goto drop;
+       }
+
+       /* We are about to mangle packet. Copy it! */
+       if (skb_cow(skb, LL_RESERVED_SPACE(rt->dst.dev)+rt->dst.header_len))
+               goto drop;
+       iph = ip_hdr(skb);
+
+       /* Decrease ttl after skb cow done */
+       ip_decrease_ttl(iph);
+
+       /*
+        *      We now generate an ICMP HOST REDIRECT giving the route
+        *      we calculated.
+        */
+       if (IPCB(skb)->flags & IPSKB_DOREDIRECT && !opt->srr &&
+           !skb_sec_path(skb))
+               ip_rt_send_redirect(skb);
+
+       skb->priority = rt_tos2priority(iph->tos);
+
+       return NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, NULL, skb,
+                      skb->dev, rt->dst.dev, ip_forward_finish);
+
+sr_failed:
+       /*
+        *      Strict routing permits no gatewaying
+        */
+        icmp_send(skb, ICMP_DEST_UNREACH, ICMP_SR_FAILED, 0);
+        goto drop;
+
+too_many_hops:
+       /* Tell the sender its packet died... */
+       IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_INHDRERRORS);
+       icmp_send(skb, ICMP_TIME_EXCEEDED, ICMP_EXC_TTL, 0);
+drop:
+       kfree_skb(skb);
+       return NET_RX_DROP;
+}