vlib: refactor checksum offload support

Type: refactor

This patch refactors the offload flags in vlib_buffer_t.
There are two main reasons behind this refactoring.
First, offload flags are insufficient to represent outer
and inner headers offloads. Second, room for these flags
in first cacheline of vlib_buffer_t is also limited.

This patch introduces a generic offload flag in first
cacheline. And detailed offload flags in 2nd cacheline
of the structure for performance optimization.

Change-Id: Icc363a142fb9208ec7113ab5bbfc8230181f6004
Signed-off-by: Mohsin Kazmi <sykazmi@cisco.com>
diff --git a/src/vnet/pg/input.c b/src/vnet/pg/input.c
index 940bf12..d3501b3 100644
--- a/src/vnet/pg/input.c
+++ b/src/vnet/pg/input.c
@@ -1537,13 +1537,14 @@
 }
 
 static_always_inline void
-fill_buffer_offload_flags (vlib_main_t * vm, u32 * buffers, u32 n_buffers,
-			   int gso_enabled, u32 gso_size)
+fill_buffer_offload_flags (vlib_main_t *vm, u32 *buffers, u32 n_buffers,
+			   u32 buffer_oflags, int gso_enabled, u32 gso_size)
 {
   for (int i = 0; i < n_buffers; i++)
     {
       vlib_buffer_t *b0 = vlib_get_buffer (vm, buffers[i]);
       u8 l4_proto = 0;
+      u32 oflags = 0;
 
       ethernet_header_t *eh =
 	(ethernet_header_t *) vlib_buffer_get_current (b0);
@@ -1574,10 +1575,11 @@
 	  vnet_buffer (b0)->l4_hdr_offset = l2hdr_sz + ip4_header_bytes (ip4);
 	  l4_proto = ip4->protocol;
 	  b0->flags |=
-	    (VNET_BUFFER_F_IS_IP4 | VNET_BUFFER_F_OFFLOAD_IP_CKSUM);
-	  b0->flags |= (VNET_BUFFER_F_L2_HDR_OFFSET_VALID
-			| VNET_BUFFER_F_L3_HDR_OFFSET_VALID |
-			VNET_BUFFER_F_L4_HDR_OFFSET_VALID);
+	    (VNET_BUFFER_F_IS_IP4 | VNET_BUFFER_F_L2_HDR_OFFSET_VALID |
+	     VNET_BUFFER_F_L3_HDR_OFFSET_VALID |
+	     VNET_BUFFER_F_L4_HDR_OFFSET_VALID);
+	  if (buffer_oflags & VNET_BUFFER_OFFLOAD_F_IP_CKSUM)
+	    oflags |= VNET_BUFFER_OFFLOAD_F_IP_CKSUM;
 	}
       else if (PREDICT_TRUE (ethertype == ETHERNET_TYPE_IP6))
 	{
@@ -1594,7 +1596,8 @@
 
       if (l4_proto == IP_PROTOCOL_TCP)
 	{
-	  b0->flags |= VNET_BUFFER_F_OFFLOAD_TCP_CKSUM;
+	  if (buffer_oflags & VNET_BUFFER_OFFLOAD_F_TCP_CKSUM)
+	    oflags |= VNET_BUFFER_OFFLOAD_F_TCP_CKSUM;
 
 	  /* only set GSO flag for chained buffers */
 	  if (gso_enabled && (b0->flags & VLIB_BUFFER_NEXT_PRESENT))
@@ -1609,8 +1612,12 @@
 	}
       else if (l4_proto == IP_PROTOCOL_UDP)
 	{
-	  b0->flags |= VNET_BUFFER_F_OFFLOAD_UDP_CKSUM;
+	  if (buffer_oflags & VNET_BUFFER_OFFLOAD_F_UDP_CKSUM)
+	    oflags |= VNET_BUFFER_OFFLOAD_F_UDP_CKSUM;
 	}
+
+      if (oflags)
+	vnet_buffer_offload_flags_set (b0, oflags);
     }
 }
 
@@ -1710,13 +1717,11 @@
 	    vnet_buffer (b)->feature_arc_index = feature_arc_index;
 	  }
 
-      if (pi->gso_enabled ||
-	  (s->buffer_flags & (VNET_BUFFER_F_OFFLOAD_TCP_CKSUM |
-			      VNET_BUFFER_F_OFFLOAD_UDP_CKSUM |
-			      VNET_BUFFER_F_OFFLOAD_IP_CKSUM)))
+      if (pi->gso_enabled || (s->buffer_flags & VNET_BUFFER_F_OFFLOAD))
 	{
 	  fill_buffer_offload_flags (vm, to_next, n_this_frame,
-				     pi->gso_enabled, pi->gso_size);
+				     s->buffer_oflags, pi->gso_enabled,
+				     pi->gso_size);
 	}
 
       n_trace = vlib_get_trace_count (vm, node);