bond: performance harvesting

- hash is great. But it is a bit too slow for the DP. Use direct array indexing
to quickly retrieve the slave interface.
- the algorithm used by flow hash is great. But it is a bit too slow for the DP.
Use l2_hash_hash() extracted from lb_hash.h which ECMP is using. It makes use
of intrinsic crc32 instruction set.
- shortcut modulo arithmetic when the operand is 2**x (where x up to 4) to
avoid division instruction.
- special case for link count == 1 in bond_tx_fn()
- use clib_mem_unaligned to access data for the packet to avoid alignment error
- Fix some typos for packet tracing.

Change-Id: I8eae3ad497061c5473aa675ba894ee0211120d25
Signed-off-by: Steven <sluong@cisco.com>

1 files changed, 15 insertions, 13 deletions

diff --git a/src/vnet/bonding/node.c b/src/vnet/bonding/node.c
index 65d3ba10470..ec251f550cf 100644
--- a/src/vnet/bonding/node.c
+++ b/src/vnet/bonding/node.c
@@ -90,20 +90,21 @@ bond_sw_if_index_rewrite (vlib_main_t * vm, vlib_node_runtime_t * node,
 
   if (PREDICT_TRUE (sif != 0))
     {
-      bif = bond_get_master_by_sw_if_index (sif->group);
+      bif = bond_get_master_by_dev_instance (sif->bif_dev_instance);
       if (PREDICT_TRUE (bif != 0))
 	{
 	  if (PREDICT_TRUE (vec_len (bif->slaves) >= 1))
 	    {
 	      if (PREDICT_TRUE (bif->admin_up == 1))
 		{
-		  if (!ethernet_frame_is_tagged (ntohs (eth->type)))
+		  ethertype = clib_mem_unaligned (&eth->type, u16);
+		  if (!ethernet_frame_is_tagged (ntohs (ethertype)))
 		    {
 		      // Let some layer2 packets pass through.
-		      if (PREDICT_TRUE ((eth->type !=
+		      if (PREDICT_TRUE ((ethertype !=
 					 htons (ETHERNET_TYPE_SLOW_PROTOCOLS))
 					&& !packet_is_cdp (eth)
-					&& (eth->type !=
+					&& (ethertype !=
 					    htons
 					    (ETHERNET_TYPE_802_1_LLDP))))
 			{
@@ -128,12 +129,13 @@ bond_sw_if_index_rewrite (vlib_main_t * vm, vlib_node_runtime_t * node,
 		    {
 		      vlan = (void *) (eth + 1);
 		      ethertype_p = &vlan->type;
-		      if (*ethertype_p == ntohs (ETHERNET_TYPE_VLAN))
+		      ethertype = clib_mem_unaligned (ethertype_p, u16);
+		      if (ethertype == ntohs (ETHERNET_TYPE_VLAN))
 			{
 			  vlan++;
 			  ethertype_p = &vlan->type;
 			}
-		      ethertype = *ethertype_p;
+		      ethertype = clib_mem_unaligned (ethertype_p, u16);
 		      if (PREDICT_TRUE ((ethertype !=
 					 htons (ETHERNET_TYPE_SLOW_PROTOCOLS))
 					&& (ethertype !=
@@ -226,10 +228,10 @@ bond_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node,
 	    b6 = vlib_get_buffer (vm, from[6]);
 	    b7 = vlib_get_buffer (vm, from[7]);
 
-	    vlib_prefetch_buffer_header (b4, STORE);
-	    vlib_prefetch_buffer_header (b5, STORE);
-	    vlib_prefetch_buffer_header (b6, STORE);
-	    vlib_prefetch_buffer_header (b7, STORE);
+	    vlib_prefetch_buffer_header (b4, LOAD);
+	    vlib_prefetch_buffer_header (b5, LOAD);
+	    vlib_prefetch_buffer_header (b6, LOAD);
+	    vlib_prefetch_buffer_header (b7, LOAD);
 
 	    CLIB_PREFETCH (b4->data, CLIB_CACHE_LINE_BYTES, LOAD);
 	    CLIB_PREFETCH (b5->data, CLIB_CACHE_LINE_BYTES, LOAD);
@@ -314,7 +316,7 @@ bond_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node,
 
 		  if (PREDICT_TRUE (n_trace > 0))
 		    {
-		      vlib_trace_buffer (vm, node, next1, b2,
+		      vlib_trace_buffer (vm, node, next2, b2,
 					 0 /* follow_chain */ );
 		      vlib_set_trace_count (vm, node, --n_trace);
 		      t0 = vlib_add_trace (vm, node, b2, sizeof (*t0));
@@ -325,7 +327,7 @@ bond_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node,
 
 		      if (PREDICT_TRUE (n_trace > 0))
 			{
-			  vlib_trace_buffer (vm, node, next1, b2,
+			  vlib_trace_buffer (vm, node, next3, b3,
 					     0 /* follow_chain */ );
 			  vlib_set_trace_count (vm, node, --n_trace);
 			  t0 = vlib_add_trace (vm, node, b3, sizeof (*t0));
@@ -358,7 +360,7 @@ bond_input_fn (vlib_main_t * vm, vlib_node_runtime_t * node,
 	      vlib_buffer_t *p2;
 
 	      p2 = vlib_get_buffer (vm, from[1]);
-	      vlib_prefetch_buffer_header (p2, STORE);
+	      vlib_prefetch_buffer_header (p2, LOAD);
 	      CLIB_PREFETCH (p2->data, CLIB_CACHE_LINE_BYTES, LOAD);
 	    }