net/mlx5: Do not hold group lock while allocating FTE in software

FTE memory allocation using alloc_fte() doesn't have any dependency
on the flow group.
Hence, do not hold flow group lock while performing alloc_fte().
This helps to reduce contention of flow group lock.

Signed-off-by: Parav Pandit <[email protected]>
Reviewed-by: Daniel Jurgens <[email protected]>
Reviewed-by: Mark Bloch <[email protected]>
Signed-off-by: Saeed Mahameed <[email protected]>

Author: paravmellanox

drivers/net/ethernet/mellanox/mlx5/core/fs_core.c

@@ -1814,24 +1814,23 @@ _mlx5_add_flow_rules(struct mlx5_flow_table *ft,
 		return rule;
 	}
 
+	fte = alloc_fte(ft, spec, flow_act);
+	if (IS_ERR(fte)) {
+		up_write_ref_node(&ft->node, false);
+		err = PTR_ERR(fte);
+		goto err_alloc_fte;
+	}
+
 	nested_down_write_ref_node(&g->node, FS_LOCK_PARENT);
 	up_write_ref_node(&ft->node, false);
 
 	err = create_auto_flow_group(ft, g);
 	if (err)
 		goto err_release_fg;
 
-	fte = alloc_fte(ft, spec, flow_act);
-	if (IS_ERR(fte)) {
-		err = PTR_ERR(fte);
-		goto err_release_fg;
-	}
-
 	err = insert_fte(g, fte);
-	if (err) {
-		kmem_cache_free(steering->ftes_cache, fte);
+	if (err)
 		goto err_release_fg;
-	}
 
 	nested_down_write_ref_node(&fte->node, FS_LOCK_CHILD);
 	up_write_ref_node(&g->node, false);
@@ -1843,6 +1842,8 @@ _mlx5_add_flow_rules(struct mlx5_flow_table *ft,
 
 err_release_fg:
 	up_write_ref_node(&g->node, false);
+	kmem_cache_free(steering->ftes_cache, fte);
+err_alloc_fte:
 	tree_put_node(&g->node, false);
 	return ERR_PTR(err);
 }