crypto: SW scheduler async crypto engine

Type: feature

This patch adds new sw_scheduler async crypto engine.
The engine transforms async frames info sync crypto ops and
delegates them to active sync engines. With the patch it
is possible to increase the single worker crypto throughput
by offloading the crypto workload to multiple workers.

By default all workers in the system will attend the crypto
workload processing. However a worker's available cycles
are limited. To avail more cycles to one worker to process
other workload (e.g. the worker core that handles the RX/TX
and IPSec stack processing), a useful cli command is added
to remove itself (or add it back later) from the heavy
crypto workload but only let other workers to process the
crypto. The command is:

 - set sw_scheduler worker <idx> crypto <on|off>

It also adds new interrupt mode to async crypto dispatch node.
This mode signals the node when new frames are enqueued
as opposed to polling mode that continuously calls dispatch node.

New cli commands:
 - set crypto async dispatch [polling|interrupt]
 - show crypto async status (displays mode and nodes' states)

Signed-off-by: PiotrX Kleski <piotrx.kleski@intel.com>
Signed-off-by: DariuszX Kazimierski <dariuszx.kazimierski@intel.com>
Reviewed-by: Fan Zhang <roy.fan.zhang@intel.com>
Change-Id: I332655f347bb9e3bc9c64166e86e393e911bdb39

1 files changed, 60 insertions, 42 deletions

diff --git a/src/vnet/crypto/node.c b/src/vnet/crypto/node.c
index 51ee63d1d62..12e6033ad80 100644
--- a/src/vnet/crypto/node.c
+++ b/src/vnet/crypto/node.c
@@ -74,60 +74,78 @@ vnet_crypto_async_add_trace (vlib_main_t * vm, vlib_node_runtime_t * node,
 static_always_inline u32
 crypto_dequeue_frame (vlib_main_t * vm, vlib_node_runtime_t * node,
 		      vnet_crypto_thread_t * ct,
-		      vnet_crypto_frame_dequeue_t * hdl,
-		      u32 n_cache, u32 * n_total)
+		      vnet_crypto_frame_dequeue_t * hdl, u32 n_cache,
+		      u32 * n_total)
 {
-  vnet_crypto_async_frame_t *cf = (hdl) (vm);
+  vnet_crypto_main_t *cm = &crypto_main;
+  u32 n_elts = 0;
+  u32 enqueue_thread_idx = ~0;
+  vnet_crypto_async_frame_t *cf = (hdl) (vm, &n_elts, &enqueue_thread_idx);
+  *n_total += n_elts;
 
-  while (cf)
+  while (cf || n_elts)
     {
-      vec_validate (ct->buffer_indice, n_cache + cf->n_elts);
-      vec_validate (ct->nexts, n_cache + cf->n_elts);
-      clib_memcpy_fast (ct->buffer_indice + n_cache, cf->buffer_indices,
-			sizeof (u32) * cf->n_elts);
-      if (cf->state == VNET_CRYPTO_FRAME_STATE_SUCCESS)
-	{
-	  clib_memcpy_fast (ct->nexts + n_cache, cf->next_node_index,
-			    sizeof (u16) * cf->n_elts);
-	}
-      else
+      if (cf)
 	{
-	  u32 i;
-	  for (i = 0; i < cf->n_elts; i++)
+	  vec_validate (ct->buffer_indice, n_cache + cf->n_elts);
+	  vec_validate (ct->nexts, n_cache + cf->n_elts);
+	  clib_memcpy_fast (ct->buffer_indice + n_cache, cf->buffer_indices,
+			    sizeof (u32) * cf->n_elts);
+	  if (cf->state == VNET_CRYPTO_FRAME_STATE_SUCCESS)
+	    {
+	      clib_memcpy_fast (ct->nexts + n_cache, cf->next_node_index,
+				sizeof (u16) * cf->n_elts);
+	    }
+	  else
 	    {
-	      if (cf->elts[i].status != VNET_CRYPTO_OP_STATUS_COMPLETED)
+	      u32 i;
+	      for (i = 0; i < cf->n_elts; i++)
 		{
-		  ct->nexts[i + n_cache] = CRYPTO_DISPATCH_NEXT_ERR_DROP;
-		  vlib_node_increment_counter (vm, node->node_index,
-					       cf->elts[i].status, 1);
+		  if (cf->elts[i].status != VNET_CRYPTO_OP_STATUS_COMPLETED)
+		    {
+		      ct->nexts[i + n_cache] = CRYPTO_DISPATCH_NEXT_ERR_DROP;
+		      vlib_node_increment_counter (vm, node->node_index,
+						   cf->elts[i].status, 1);
+		    }
+		  else
+		    ct->nexts[i + n_cache] = cf->next_node_index[i];
 		}
-	      else
-		ct->nexts[i + n_cache] = cf->next_node_index[i];
 	    }
-	}
-      n_cache += cf->n_elts;
-      *n_total += cf->n_elts;
-      if (n_cache >= VLIB_FRAME_SIZE)
-	{
-	  vlib_buffer_enqueue_to_next (vm, node, ct->buffer_indice, ct->nexts,
-				       n_cache);
-	  n_cache = 0;
-	}
-
-      if (PREDICT_FALSE (node->flags & VLIB_NODE_FLAG_TRACE))
-	{
-	  u32 i;
+	  n_cache += cf->n_elts;
+	  if (n_cache >= VLIB_FRAME_SIZE)
+	    {
+	      vlib_buffer_enqueue_to_next (vm, node, ct->buffer_indice,
+					   ct->nexts, n_cache);
+	      n_cache = 0;
+	    }
 
-	  for (i = 0; i < cf->n_elts; i++)
+	  if (PREDICT_FALSE (node->flags & VLIB_NODE_FLAG_TRACE))
 	    {
-	      vlib_buffer_t *b = vlib_get_buffer (vm, cf->buffer_indices[i]);
-	      if (b->flags & VLIB_BUFFER_IS_TRACED)
-		vnet_crypto_async_add_trace (vm, node, b, cf->op,
-					     cf->elts[i].status);
+	      u32 i;
+
+	      for (i = 0; i < cf->n_elts; i++)
+		{
+		  vlib_buffer_t *b = vlib_get_buffer (vm,
+						      cf->buffer_indices[i]);
+		  if (b->flags & VLIB_BUFFER_IS_TRACED)
+		    vnet_crypto_async_add_trace (vm, node, b, cf->op,
+						 cf->elts[i].status);
+		}
 	    }
+	  vnet_crypto_async_free_frame (vm, cf);
+	}
+      /* signal enqueue-thread to dequeue the processed frame (n_elts>0) */
+      if (cm->dispatch_mode == VNET_CRYPTO_ASYNC_DISPATCH_INTERRUPT
+	  && n_elts > 0)
+	{
+	  vlib_node_set_interrupt_pending (vlib_mains[enqueue_thread_idx],
+					   cm->crypto_node_index);
 	}
-      vnet_crypto_async_free_frame (vm, cf);
-      cf = (hdl) (vm);
+
+      n_elts = 0;
+      enqueue_thread_idx = 0;
+      cf = (hdl) (vm, &n_elts, &enqueue_thread_idx);
+      *n_total += n_elts;
     }
 
   return n_cache;