perfmon: collect data on selected thread(s)

Add missing pre-input node runtime fork and refork code.
unix-epoll-input runs on all threads; each instance needs its own
runtime stats.

Change-Id: I16b02e42d0c95f863161176c4bb9f9917bef809d
Signed-off-by: Dave Barach <dave@barachs.net>

2 files changed, 43 insertions, 9 deletions

diff --git a/src/vlib/main.c b/src/vlib/main.c
index 0e480fabe2a..3048a0d72de 100644
--- a/src/vlib/main.c
+++ b/src/vlib/main.c
@@ -1105,8 +1105,6 @@ dispatch_pcap_trace (vlib_main_t * vm,
     }
 }
 
-u64 oingo0, oingo1;
-
 static_always_inline u64
 dispatch_node (vlib_main_t * vm,
 	       vlib_node_runtime_t * node,
@@ -1207,12 +1205,6 @@ dispatch_node (vlib_main_t * vm,
       vm->main_loop_vectors_processed += n;
       vm->main_loop_nodes_processed += n > 0;
 
-      if (pmc_delta[0] || pmc_delta[1])
-	{
-	  oingo0 += pmc_delta[0];
-	  oingo1 += pmc_delta[1];
-	}
-
       v = vlib_node_runtime_update_stats (vm, node,
 					  /* n_calls */ 1,
 					  /* n_vectors */ n,
diff --git a/src/vlib/threads.c b/src/vlib/threads.c
index 45e4d89b7ab..e6ac6db543a 100644
--- a/src/vlib/threads.c
+++ b/src/vlib/threads.c
@@ -801,7 +801,7 @@ start_workers (vlib_main_t * vm)
 
 	      /* fork the frame dispatch queue */
 	      nm_clone->pending_frames = 0;
-	      vec_validate (nm_clone->pending_frames, 10);	/* $$$$$?????? */
+	      vec_validate (nm_clone->pending_frames, 10);
 	      _vec_len (nm_clone->pending_frames) = 0;
 
 	      /* fork nodes */
@@ -850,6 +850,21 @@ start_workers (vlib_main_t * vm)
 					 n->runtime_data_bytes));
 	      }
 
+	      nm_clone->nodes_by_type[VLIB_NODE_TYPE_PRE_INPUT] =
+		vec_dup_aligned (nm->nodes_by_type[VLIB_NODE_TYPE_PRE_INPUT],
+				 CLIB_CACHE_LINE_BYTES);
+	      vec_foreach (rt,
+			   nm_clone->nodes_by_type[VLIB_NODE_TYPE_PRE_INPUT])
+	      {
+		vlib_node_t *n = vlib_get_node (vm, rt->node_index);
+		rt->thread_index = vm_clone->thread_index;
+		/* copy initial runtime_data from node */
+		if (n->runtime_data && n->runtime_data_bytes > 0)
+		  clib_memcpy (rt->runtime_data, n->runtime_data,
+			       clib_min (VLIB_NODE_RUNTIME_DATA_SIZE,
+					 n->runtime_data_bytes));
+	      }
+
 	      nm_clone->processes = vec_dup_aligned (nm->processes,
 						     CLIB_CACHE_LINE_BYTES);
 
@@ -1173,6 +1188,33 @@ vlib_worker_thread_node_refork (void)
 
   vec_free (old_rt);
 
+  /* re-clone pre-input nodes */
+  old_rt = nm_clone->nodes_by_type[VLIB_NODE_TYPE_PRE_INPUT];
+  nm_clone->nodes_by_type[VLIB_NODE_TYPE_PRE_INPUT] =
+    vec_dup_aligned (nm->nodes_by_type[VLIB_NODE_TYPE_PRE_INPUT],
+		     CLIB_CACHE_LINE_BYTES);
+
+  vec_foreach (rt, nm_clone->nodes_by_type[VLIB_NODE_TYPE_PRE_INPUT])
+  {
+    vlib_node_t *n = vlib_get_node (vm, rt->node_index);
+    rt->thread_index = vm_clone->thread_index;
+    /* copy runtime_data, will be overwritten later for existing rt */
+    if (n->runtime_data && n->runtime_data_bytes > 0)
+      clib_memcpy_fast (rt->runtime_data, n->runtime_data,
+			clib_min (VLIB_NODE_RUNTIME_DATA_SIZE,
+				  n->runtime_data_bytes));
+  }
+
+  for (j = 0; j < vec_len (old_rt); j++)
+    {
+      rt = vlib_node_get_runtime (vm_clone, old_rt[j].node_index);
+      rt->state = old_rt[j].state;
+      clib_memcpy_fast (rt->runtime_data, old_rt[j].runtime_data,
+			VLIB_NODE_RUNTIME_DATA_SIZE);
+    }
+
+  vec_free (old_rt);
+
   nm_clone->processes = vec_dup_aligned (nm->processes,
 					 CLIB_CACHE_LINE_BYTES);
 }