Add vlib_buffer_copy_indices inline function

This reverts commit 1e59f9ddbdda14591967e1d66eab8623f9ba58e4.

Change-Id: Iae1d372b887e170d28cac2fe4c61325ee5a5894a
Signed-off-by: Damjan Marion <damarion@cisco.com>

6 files changed, 20 insertions, 15 deletions

diff --git a/src/plugins/avf/input.c b/src/plugins/avf/input.c
index b784bf731c1..fc884ce39f6 100644
--- a/src/plugins/avf/input.c
+++ b/src/plugins/avf/input.c
@@ -295,7 +295,7 @@ avf_device_input_inline (vlib_main_t * vm, vlib_node_runtime_t * node,
 
       or_q1x4 |= q1x4;
       u64x4_store_unaligned (q1x4, ptd->qw1s + n_rx_packets);
-      clib_memcpy_fast (bi, rxq->bufs + next, 4 * sizeof (u32));
+      vlib_buffer_copy_indices (bi, rxq->bufs + next, 4);
 
       /* next */
       next = (next + 4) & mask;
diff --git a/src/plugins/avf/output.c b/src/plugins/avf/output.c
index 0db333be823..47803fc2729 100644
--- a/src/plugins/avf/output.c
+++ b/src/plugins/avf/output.c
@@ -66,7 +66,7 @@ avf_tx_enqueue (vlib_main_t * vm, avf_txq_t * txq, u32 * buffers,
       if (or_flags & VLIB_BUFFER_NEXT_PRESENT)
 	goto one_by_one;
 
-      clib_memcpy_fast (txq->bufs + next, buffers, sizeof (u32) * 4);
+      vlib_buffer_copy_indices (txq->bufs + next, buffers, 4);
 
       if (use_va_dma)
 	{
diff --git a/src/plugins/marvell/pp2/output.c b/src/plugins/marvell/pp2/output.c
index dbd106a5930..911b2f55a17 100644
--- a/src/plugins/marvell/pp2/output.c
+++ b/src/plugins/marvell/pp2/output.c
@@ -101,7 +101,7 @@ mrvl_pp2_interface_tx (vlib_main_t * vm,
       buffers = vlib_frame_vector_args (frame);
       u16 n_copy = clib_min (outq->size - slot, n_sent);
 
-      clib_memcpy_fast (outq->buffers + slot, buffers, n_copy * sizeof (u32));
+      vlib_buffer_copy_indices (outq->buffers + slot, buffers, n_copy);
       if (PREDICT_FALSE (n_copy < n_sent))
 	clib_memcpy_fast (outq->buffers, buffers + n_copy,
 			  (n_sent - n_copy) * sizeof (u32));
diff --git a/src/vlib/buffer_funcs.h b/src/vlib/buffer_funcs.h
index 8fbb58d68b3..f5781c8b173 100644
--- a/src/vlib/buffer_funcs.h
+++ b/src/vlib/buffer_funcs.h
@@ -65,6 +65,12 @@ vlib_get_buffer (vlib_main_t * vm, u32 buffer_index)
 }
 
 static_always_inline void
+vlib_buffer_copy_indices (u32 * dst, u32 * src, u32 n_indices)
+{
+  clib_memcpy_fast (dst, src, n_indices * sizeof (u32));
+}
+
+static_always_inline void
 vlib_buffer_copy_template (vlib_buffer_t * b, vlib_buffer_t * bt)
 {
   clib_memcpy_fast (b, bt, STRUCT_OFFSET_OF (vlib_buffer_t, template_end));
@@ -454,7 +460,7 @@ vlib_buffer_alloc_from_free_list (vlib_main_t * vm,
       /* following code is intentionaly duplicated to allow compiler
          to optimize fast path when n_buffers is constant value */
       src = fl->buffers + len - n_buffers;
-      clib_memcpy_fast (buffers, src, n_buffers * sizeof (u32));
+      vlib_buffer_copy_indices (buffers, src, n_buffers);
       _vec_len (fl->buffers) -= n_buffers;
 
       /* Verify that buffers are known free. */
@@ -465,7 +471,7 @@ vlib_buffer_alloc_from_free_list (vlib_main_t * vm,
     }
 
   src = fl->buffers + len - n_buffers;
-  clib_memcpy_fast (buffers, src, n_buffers * sizeof (u32));
+  vlib_buffer_copy_indices (buffers, src, n_buffers);
   _vec_len (fl->buffers) -= n_buffers;
 
   /* Verify that buffers are known free. */
diff --git a/src/vlib/buffer_node.h b/src/vlib/buffer_node.h
index 2163b72de5e..7e488783fc5 100644
--- a/src/vlib/buffer_node.h
+++ b/src/vlib/buffer_node.h
@@ -383,7 +383,7 @@ vlib_buffer_enqueue_to_next (vlib_main_t * vm, vlib_node_runtime_t * node,
 #ifdef CLIB_HAVE_VEC512
       if (n_enqueued >= 32)
 	{
-	  clib_memcpy_fast (to_next, buffers, 32 * sizeof (u32));
+	  vlib_buffer_copy_indices (to_next, buffers, 32);
 	  nexts += 32;
 	  to_next += 32;
 	  buffers += 32;
@@ -397,7 +397,7 @@ vlib_buffer_enqueue_to_next (vlib_main_t * vm, vlib_node_runtime_t * node,
 #ifdef CLIB_HAVE_VEC256
       if (n_enqueued >= 16)
 	{
-	  clib_memcpy_fast (to_next, buffers, 16 * sizeof (u32));
+	  vlib_buffer_copy_indices (to_next, buffers, 16);
 	  nexts += 16;
 	  to_next += 16;
 	  buffers += 16;
@@ -411,7 +411,7 @@ vlib_buffer_enqueue_to_next (vlib_main_t * vm, vlib_node_runtime_t * node,
 #ifdef CLIB_HAVE_VEC128
       if (n_enqueued >= 8)
 	{
-	  clib_memcpy_fast (to_next, buffers, 8 * sizeof (u32));
+	  vlib_buffer_copy_indices (to_next, buffers, 8);
 	  nexts += 8;
 	  to_next += 8;
 	  buffers += 8;
@@ -424,7 +424,7 @@ vlib_buffer_enqueue_to_next (vlib_main_t * vm, vlib_node_runtime_t * node,
 
       if (n_enqueued >= 4)
 	{
-	  clib_memcpy_fast (to_next, buffers, 4 * sizeof (u32));
+	  vlib_buffer_copy_indices (to_next, buffers, 4);
 	  nexts += 4;
 	  to_next += 4;
 	  buffers += 4;
@@ -459,7 +459,7 @@ vlib_buffer_enqueue_to_single_next (vlib_main_t * vm,
 
   if (PREDICT_TRUE (n_left_to_next >= count))
     {
-      clib_memcpy_fast (to_next, buffers, count * sizeof (u32));
+      vlib_buffer_copy_indices (to_next, buffers, count);
       n_left_to_next -= count;
       vlib_put_next_frame (vm, node, next_index, n_left_to_next);
       return;
@@ -467,7 +467,7 @@ vlib_buffer_enqueue_to_single_next (vlib_main_t * vm,
 
   n_enq = n_left_to_next;
 next:
-  clib_memcpy_fast (to_next, buffers, n_enq * sizeof (u32));
+  vlib_buffer_copy_indices (to_next, buffers, n_enq);
   n_left_to_next -= n_enq;
 
   if (PREDICT_FALSE (count > n_enq))
diff --git a/src/vnet/pg/input.c b/src/vnet/pg/input.c
index 792c67cd88d..6088b91271e 100644
--- a/src/vnet/pg/input.c
+++ b/src/vnet/pg/input.c
@@ -1590,13 +1590,12 @@ pg_generate_packets (vlib_node_runtime_t * node,
       head = clib_fifo_head (bi0->buffer_fifo);
 
       if (head + n_this_frame <= end)
-	clib_memcpy_fast (to_next, head, n_this_frame * sizeof (u32));
+	vlib_buffer_copy_indices (to_next, head, n_this_frame);
       else
 	{
 	  u32 n = end - head;
-	  clib_memcpy_fast (to_next + 0, head, n * sizeof (u32));
-	  clib_memcpy_fast (to_next + n, start,
-			    (n_this_frame - n) * sizeof (u32));
+	  vlib_buffer_copy_indices (to_next + 0, head, n);
+	  vlib_buffer_copy_indices (to_next + n, start, n_this_frame - n);
 	}
 
       if (s->replay_packet_templates == 0)