VPP-1277: IPIP - Copy TOS/TC from inner packet to outer.

Add support for either copying TOS/TC from inner packet to outer,
or set to fixed value.

Change-Id: I716a95f875349acec94317b266c8cf9f2f81a785
Signed-off-by: Ole Troan <ot@cisco.com>

5 files changed, 109 insertions, 18 deletions

diff --git a/src/vnet/ipip/ipip.api b/src/vnet/ipip/ipip.api
index 988eee599b1..95fc48a5962 100644
--- a/src/vnet/ipip/ipip.api
+++ b/src/vnet/ipip/ipip.api
@@ -13,17 +13,46 @@
  * limitations under the License.
  */
 
-option version = "1.0.0";
+/**
+ * The IPIP module implements IP{v4,v6} over IP{v4,v6} tunnelling as
+ * described in RFC2473 and to some extent the largely historical
+ * RFC1853.  The module also supports an IPv4 over IPv6 automatic
+ * tunnelling mechanism called 6RD (RFC5969).
+ *
+ * The IPIP API module supports a CRD model for adding, deleting and
+ * listing tunnels. A tunnel is represented as an interface in
+ * VPP. The "handle" representing a tunnel is the sw_if_index.  As any
+ * interface, the user must configure an IPv4 and/or IPv6 address on
+ * the interface. This is the inner or payload protocol.
+ *
+ * Tunnel MTU: The tunnel MTU (the payload MTU) is configurable per
+ * protocol. If a tunnel MTU is larger than the path MTU, the outer
+ * packet will be fragmented. Fragmentation support is configurable,
+ * as it can have severe performance issues, and might be used as an
+ * attack vector (the remote side must reassemble.)
+ *
+ * Traffic class / TOS field can either be configured to a fixed
+ * value, or can be copied from the inner to the outer header.
+ * (For now we have stolen ~0 to indicate copy).
+ *
+ * Note:
+ *
+ * - The Tunnel encapsulation limit described in RFC2473 is not
+ *   implemented.
+ *
+ * - ICMP proxying, as in a tunnel head-end receiving ICMP erors on
+ *   the outer packet is currently not relayed to the original source
+ *   of the packet.
+ *
+ * - PMTUD / MTU probing and tunnel keepalives are not yet implemented.
+ *
+ */
+
+option version = "1.1.0";
 
-/** \brief Create or delete an IPIP tunnel
-    @param client_index - opaque cookie to identify the sender
-    @param context - sender context, to match reply w/ request
-    @param is_ipv6 - Use 0 for IPv4, 1 for IPv6
-    @param instance - optional unique custom device instance, else ~0.
-    @param src_address - Source IP address
-    @param dst_address - Destination IP address, can be multicast
-    @param fib_index - Encap FIB table ID
-*/
+/**
+ * Create an IP{v4,v6} over IP{v4,v6} tunnel.
+ */
 define ipip_add_tunnel
 {
   u32 client_index;
@@ -33,6 +62,8 @@ define ipip_add_tunnel
   u8 src_address[16];
   u8 dst_address[16];
   u32 fib_index;
+  u8 tc_tos;		/* If ~0, the TOS/TC value is copied from
+			   inner packet, otherwise set to value */
 };
 
 define ipip_add_tunnel_reply
@@ -42,6 +73,9 @@ define ipip_add_tunnel_reply
   u32 sw_if_index;
 };
 
+/**
+ * Delete an IP{v4,v6} over IP{v4,v6} tunnel.
+ */
 autoreply define ipip_del_tunnel
 {
   u32 client_index;
@@ -49,6 +83,9 @@ autoreply define ipip_del_tunnel
   u32 sw_if_index;
 };
 
+/**
+ * Create an IPv4 over IPv6 automatic tunnel (6RD)
+ */
 define ipip_6rd_add_tunnel
 {
   u32 client_index;
@@ -60,6 +97,8 @@ define ipip_6rd_add_tunnel
   u8 ip6_prefix_len;
   u8 ip4_prefix_len;
   u8 security_check;
+  u8 tc_tos;		/* If ~0, the TOS/TC value is copied from
+			   inner packet, otherwise set to value */
 };
 
 define ipip_6rd_add_tunnel_reply
@@ -69,6 +108,9 @@ define ipip_6rd_add_tunnel_reply
   u32 sw_if_index;
 };
 
+/**
+ * Delete an IPv4 over IPv6 automatic tunnel (6RD)
+ */
 autoreply define ipip_6rd_del_tunnel
 {
   u32 client_index;
@@ -76,6 +118,9 @@ autoreply define ipip_6rd_del_tunnel
   u32 sw_if_index;
 };
 
+/**
+ * List all IPIP tunnels
+ */
 define ipip_tunnel_dump
 {
   u32 client_index;
@@ -92,6 +137,7 @@ define ipip_tunnel_details
   u8 src_address[16];
   u8 dst_address[16];
   u32 fib_index;
+  u8 tc_tos;
 };
 
 /*
diff --git a/src/vnet/ipip/ipip.c b/src/vnet/ipip/ipip.c
index 82c961cdddd..c39d27d5e4f 100644
--- a/src/vnet/ipip/ipip.c
+++ b/src/vnet/ipip/ipip.c
@@ -74,6 +74,8 @@ ipip_build_rewrite (vnet_main_t * vnm, u32 sw_if_index,
       ip4->src_address.as_u32 = t->tunnel_src.ip4.as_u32;
       ip4->dst_address.as_u32 = t->tunnel_dst.ip4.as_u32;
       ip4->checksum = ip4_header_checksum (ip4);
+      if (t->tc_tos != 0xFF)
+	ip4->tos = t->tc_tos;
       break;
 
     case IPIP_TRANSPORT_IP6:
@@ -81,6 +83,8 @@ ipip_build_rewrite (vnet_main_t * vnm, u32 sw_if_index,
       ip6 = (ip6_header_t *) rewrite;
       ip6->ip_version_traffic_class_and_flow_label =
 	clib_host_to_net_u32 (6 << 28);
+      if (t->tc_tos != 0xFF)
+	ip6_set_traffic_class_network_order (ip6, t->tc_tos);
       ip6->hop_limit = 64;
       /* fixup ip6 header length and protocol after-the-fact */
       ip6->src_address.as_u64[0] = t->tunnel_src.ip6.as_u64[0];
@@ -88,6 +92,7 @@ ipip_build_rewrite (vnet_main_t * vnm, u32 sw_if_index,
       ip6->dst_address.as_u64[0] = t->tunnel_dst.ip6.as_u64[0];
       ip6->dst_address.as_u64[1] = t->tunnel_dst.ip6.as_u64[1];
       break;
+
     default:
       /* pass through */
       ;
@@ -100,11 +105,29 @@ ipip4_fixup (vlib_main_t * vm, ip_adjacency_t * adj, vlib_buffer_t * b,
 	     const void *data)
 {
   ip4_header_t *ip4;
+  const ipip_tunnel_t *t = data;
 
   ip4 = vlib_buffer_get_current (b);
   ip4->length = clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b));
-  ip4->protocol =
-    adj->ia_link == VNET_LINK_IP6 ? IP_PROTOCOL_IPV6 : IP_PROTOCOL_IP_IN_IP;
+  switch (adj->ia_link)
+    {
+    case VNET_LINK_IP6:
+      ip4->protocol = IP_PROTOCOL_IPV6;
+      if (t->tc_tos == 0xFF)
+	ip4->tos =
+	  ip6_traffic_class_network_order ((const ip6_header_t *) (ip4 + 1));
+      break;
+
+    case VNET_LINK_IP4:
+      ip4->protocol = IP_PROTOCOL_IP_IN_IP;
+      if (t->tc_tos == 0xFF)
+	ip4->tos = ((ip4_header_t *) (ip4 + 1))->tos;
+      break;
+
+    default:
+      break;
+    }
+
   ip4->checksum = ip4_header_checksum (ip4);
 }
 
@@ -113,13 +136,32 @@ ipip6_fixup (vlib_main_t * vm, ip_adjacency_t * adj, vlib_buffer_t * b,
 	     const void *data)
 {
   ip6_header_t *ip6;
+  const ipip_tunnel_t *t = data;
 
   ip6 = vlib_buffer_get_current (b);
   ip6->payload_length =
     clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b) -
 			  sizeof (*ip6));
-  ip6->protocol =
-    adj->ia_link == VNET_LINK_IP6 ? IP_PROTOCOL_IPV6 : IP_PROTOCOL_IP_IN_IP;
+  switch (adj->ia_link)
+    {
+    case VNET_LINK_IP6:
+      ip6->protocol = IP_PROTOCOL_IPV6;
+      if (t->tc_tos == 0xFF)
+	ip6_set_traffic_class_network_order (ip6,
+					     ip6_traffic_class_network_order ((const ip6_header_t *) (ip6 + 1)));
+      break;
+
+    case VNET_LINK_IP4:
+      ip6->protocol = IP_PROTOCOL_IP_IN_IP;
+      if (t->tc_tos == 0xFF)
+	ip6_set_traffic_class_network_order (ip6,
+					     ((ip4_header_t *) (ip6 +
+								1))->tos);
+      break;
+
+    default:
+      break;
+    }
 }
 
 static void
@@ -216,7 +258,7 @@ ipip_update_adj (vnet_main_t * vnm, u32 sw_if_index, adj_index_t ai)
 
   f = t->transport == IPIP_TRANSPORT_IP6 ? ipip6_fixup : ipip4_fixup;
 
-  adj_nbr_midchain_update_rewrite (ai, f, NULL,
+  adj_nbr_midchain_update_rewrite (ai, f, t,
 				   (VNET_LINK_ETHERNET ==
 				    adj_get_link_type (ai) ?
 				    ADJ_FLAG_MIDCHAIN_NO_COUNT :
@@ -420,7 +462,7 @@ ipip_fib_delete (ipip_tunnel_t * t)
 int
 ipip_add_tunnel (ipip_transport_t transport,
 		 u32 instance, ip46_address_t * src, ip46_address_t * dst,
-		 u32 fib_index, u32 * sw_if_indexp)
+		 u32 fib_index, u8 tc_tos, u32 * sw_if_indexp)
 {
   ipip_main_t *gm = &ipip_main;
   vnet_main_t *vnm = gm->vnet_main;
@@ -467,6 +509,7 @@ ipip_add_tunnel (ipip_transport_t transport,
   t->hw_if_index = hw_if_index;
   t->fib_index = fib_index;
   t->sw_if_index = sw_if_index;
+  t->tc_tos = tc_tos;
 
   t->transport = transport;
   vec_validate_init_empty (gm->tunnel_index_by_sw_if_index, sw_if_index, ~0);
diff --git a/src/vnet/ipip/ipip.h b/src/vnet/ipip/ipip.h
index f52094f1f0c..6afb188f8ee 100644
--- a/src/vnet/ipip/ipip.h
+++ b/src/vnet/ipip/ipip.h
@@ -81,6 +81,7 @@ typedef struct
   u32 sw_if_index;
   u32 dev_instance;		/* Real device instance in tunnel vector */
   u32 user_instance;		/* Instance name being shown to user */
+  u8 tc_tos;
 
   union
   {
@@ -148,7 +149,7 @@ sixrd_get_addr_net (const ipip_tunnel_t * t, u64 dal)
 
 int ipip_add_tunnel (ipip_transport_t transport, u32 instance,
 		     ip46_address_t * src, ip46_address_t * dst,
-		     u32 fib_index, u32 * sw_if_indexp);
+		     u32 fib_index, u8 tc_tos, u32 * sw_if_indexp);
 int ipip_del_tunnel (u32 sw_if_index);
 int sixrd_add_tunnel (ip6_address_t * ip6_prefix, u8 ip6_prefix_len,
 		      ip4_address_t * ip4_prefix, u8 ip4_prefix_len,
diff --git a/src/vnet/ipip/ipip_api.c b/src/vnet/ipip/ipip_api.c
index a6633296515..455792ba133 100644
--- a/src/vnet/ipip/ipip_api.c
+++ b/src/vnet/ipip/ipip_api.c
@@ -69,7 +69,7 @@ vl_api_ipip_add_tunnel_t_handler (vl_api_ipip_add_tunnel_t * mp)
 
   rv = ipip_add_tunnel (mp->is_ipv6 ? IPIP_TRANSPORT_IP6 : IPIP_TRANSPORT_IP4,
 			ntohl (mp->instance), &src, &dst,
-			ntohl (mp->fib_index), &sw_if_index);
+			ntohl (mp->fib_index), mp->tc_tos, &sw_if_index);
 
   /* *INDENT-OFF* */
   REPLY_MACRO2(VL_API_IPIP_ADD_TUNNEL_REPLY,
diff --git a/src/vnet/ipip/ipip_cli.c b/src/vnet/ipip/ipip_cli.c
index 45e6451d69a..7a68c20319e 100644
--- a/src/vnet/ipip/ipip_cli.c
+++ b/src/vnet/ipip/ipip_cli.c
@@ -72,6 +72,7 @@ static clib_error_t *create_ipip_tunnel_command_fn(vlib_main_t *vm,
 		       &src,
 		       &dst,
 		       fib_index,
+		       0,
 		       &sw_if_index);
 
   switch (rv) {