bond: tx performance enhancement part deux

- Reduce per packet cost by buffering the output packet buffer indexes in the queue and
process the queue outside the packet processing loop.
- Move unnecessary variable initialization outside of the while loop.
- There is no need to save the old interface if tracing is not enabled.

Test result for 256 bytes packet comparison. Other packet size shows similar improvement.

With the patch
--------------
BondEthernet0-output             active              52836        13526016               0          1.71e1          256.00
BondEthernet0-tx                 active              52836        13526016               0          2.68e1          256.00
TenGigabitEthernet6/0/0-output   active              52836         6762896               0          9.17e0          127.99
TenGigabitEthernet6/0/0-tx       active              52836         6762896               0          6.97e1          127.99
TenGigabitEthernet6/0/1-output   active              52836         6763120               0          9.40e0          128.00
TenGigabitEthernet6/0/1-tx       active              52836         6763120               0          7.00e1          128.00
bond-input                       active              52836        13526016               0          1.76e1          256.00

Without the patch
-----------------
BondEthernet0-output             active              60858        15579648               0          1.73e1          256.00
BondEthernet0-tx                 active              60858        15579648               0          2.94e1          256.00
TenGigabitEthernet6/0/0-output   active              60858         7789626               0          9.29e0          127.99
TenGigabitEthernet6/0/0-tx       active              60858         7789626               0          7.01e1          127.99
TenGigabitEthernet6/0/1-output   active              60858         7790022               0          9.31e0          128.00
TenGigabitEthernet6/0/1-tx       active              60858         7790022               0          7.10e1          128.00
bond-input                       active              60858        15579648               0          1.77e1          256.00

Change-Id: Ib6d73a63ceeaa2f1397ceaf4c5391c57fd865b04
Signed-off-by: Steven <sluong@cisco.com>

src/vnet/bonding/cli.c

@@ -198,7 +198,6 @@ bond_delete_if (vlib_main_t * vm, u32 sw_if_index)
   slave_if_t *sif;
   vnet_hw_interface_t *hw;
   u32 *sif_sw_if_index;
-  u32 thread_index;
   u32 **s_list = 0;
   u32 i;
 
@@ -232,12 +231,6 @@ bond_delete_if (vlib_main_t * vm, u32 sw_if_index)
 
   clib_bitmap_free (bif->port_number_bitmap);
   hash_unset (bm->bond_by_sw_if_index, bif->sw_if_index);
-  for (thread_index = 0; thread_index < vlib_get_thread_main ()->n_vlib_mains;
-       thread_index++)
-    {
-      vec_free (bif->per_thread_info[thread_index].frame);
-    }
-  vec_free (bif->per_thread_info);
   memset (bif, 0, sizeof (*bif));
   pool_put (bm->interfaces, bif);
 
@@ -310,9 +303,6 @@ bond_create_if (vlib_main_t * vm, bond_create_if_args_t * args)
   sw = vnet_get_hw_sw_interface (vnm, bif->hw_if_index);
   bif->sw_if_index = sw->sw_if_index;
   bif->group = bif->sw_if_index;
-  vec_validate_aligned (bif->per_thread_info,
-			vlib_get_thread_main ()->n_vlib_mains - 1,
-			CLIB_CACHE_LINE_BYTES);
   if (vlib_get_thread_main ()->n_vlib_mains > 1)
     clib_spinlock_init (&bif->lockp);
 
@@ -431,6 +421,8 @@ bond_enslave (vlib_main_t * vm, bond_enslave_args_t * args)
   vnet_interface_main_t *im = &vnm->interface_main;
   vnet_hw_interface_t *bif_hw, *sif_hw;
   vnet_sw_interface_t *sw;
+  u32 thread_index;
+  u32 sif_if_index;
 
   bif = bond_get_master_by_sw_if_index (args->group);
   if (!bif)
@@ -527,6 +519,20 @@ bond_enslave (vlib_main_t * vm, bond_enslave_args_t * args)
       bond_enable_collecting_distributing (vm, sif);
     }
 
+  vec_foreach_index (thread_index, bm->per_thread_data)
+  {
+    bond_per_thread_data_t *ptd = vec_elt_at_index (bm->per_thread_data,
+						    thread_index);
+
+    vec_validate_aligned (ptd->per_port_queue, sif->sw_if_index,
+			  CLIB_CACHE_LINE_BYTES);
+
+    vec_foreach_index (sif_if_index, ptd->per_port_queue)
+    {
+      ptd->per_port_queue[sif_if_index].n_buffers = 0;
+    }
+  }
+
   args->rv = vnet_feature_enable_disable ("device-input", "bond-input",
 					  sif_hw->hw_if_index, 1, 0, 0);
 
@@ -755,6 +761,9 @@ bond_cli_init (vlib_main_t * vm)
   bm->vlib_main = vm;
   bm->vnet_main = vnet_get_main ();
   vec_validate_aligned (bm->slave_by_sw_if_index, 1, CLIB_CACHE_LINE_BYTES);
+  vec_validate_aligned (bm->per_thread_data,
+			vlib_get_thread_main ()->n_vlib_mains - 1,
+			CLIB_CACHE_LINE_BYTES);
 
   return 0;
 }

src/vnet/bonding/node.h

@@ -139,9 +139,15 @@ typedef CLIB_PACKED (struct
 
 typedef struct
 {
-  vlib_frame_t **frame;
+  CLIB_CACHE_LINE_ALIGN_MARK (cacheline0);
+  u32 buffers[VLIB_FRAME_SIZE];
+  u32 n_buffers;
+} bond_per_port_queue_t;
 
-} bond_if_per_thread_t;
+typedef struct
+{
+  bond_per_port_queue_t *per_port_queue;
+} bond_per_thread_data_t;
 
 typedef struct
 {
@@ -175,7 +181,6 @@ typedef struct
   u8 hw_address[6];
 
   clib_spinlock_t lockp;
-  bond_if_per_thread_t *per_thread_info;
 } bond_if_t;
 
 typedef struct
@@ -292,7 +297,7 @@ typedef struct
   /* pool of bonding interfaces */
   bond_if_t *interfaces;
 
-  /* pool of lacp neighbors */
+  /* pool of slave interfaces */
   slave_if_t *neighbors;
 
   /* rapidly find a bond by vlib software interface index */
@@ -308,6 +313,8 @@ typedef struct
   lacp_enable_disable_func lacp_enable_disable;
 
   uword *slave_by_sw_if_index;
+
+  bond_per_thread_data_t *per_thread_data;
 } bond_main_t;
 
 /* bond packet trace capture */