diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index bee207ce589a266c990eb2b40f13e8d8de05d654..3d8c52220f1ff2ec774e9f64b347b38c46ef6262 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -122,8 +122,8 @@ static void init_rem_hits(void)
 }
 
 static size_t
-callchain__fprintf_graph(FILE *fp, struct callchain_node *self,
-			u64 total_samples, int depth, int depth_mask)
+__callchain__fprintf_graph(FILE *fp, struct callchain_node *self,
+			   u64 total_samples, int depth, int depth_mask)
 {
 	struct rb_node *node, *next;
 	struct callchain_node *child;
@@ -174,9 +174,9 @@ callchain__fprintf_graph(FILE *fp, struct callchain_node *self,
 						      new_total,
 						      cumul);
 		}
-		ret += callchain__fprintf_graph(fp, child, new_total,
-						depth + 1,
-						new_depth_mask | (1 << depth));
+		ret += __callchain__fprintf_graph(fp, child, new_total,
+						  depth + 1,
+						  new_depth_mask | (1 << depth));
 		node = next;
 	}
 
@@ -196,6 +196,33 @@ callchain__fprintf_graph(FILE *fp, struct callchain_node *self,
 	return ret;
 }
 
+static size_t
+callchain__fprintf_graph(FILE *fp, struct callchain_node *self,
+			 u64 total_samples)
+{
+	struct callchain_list *chain;
+	int i = 0;
+	int ret = 0;
+
+	list_for_each_entry(chain, &self->val, list) {
+		if (chain->ip >= PERF_CONTEXT_MAX)
+			continue;
+
+		if (!i++ && sort_by_sym_first)
+			continue;
+
+		if (chain->sym)
+			ret += fprintf(fp, "                %s\n", chain->sym->name);
+		else
+			ret += fprintf(fp, "                %p\n",
+					(void *)(long)chain->ip);
+	}
+
+	ret += __callchain__fprintf_graph(fp, self, total_samples, 1, 1);
+
+	return ret;
+}
+
 static size_t
 callchain__fprintf_flat(FILE *fp, struct callchain_node *self,
 			u64 total_samples)
@@ -244,8 +271,7 @@ hist_entry_callchain__fprintf(FILE *fp, struct hist_entry *self,
 			break;
 		case CHAIN_GRAPH_ABS: /* Falldown */
 		case CHAIN_GRAPH_REL:
-			ret += callchain__fprintf_graph(fp, chain,
-							total_samples, 1, 1);
+			ret += callchain__fprintf_graph(fp, chain, total_samples);
 		case CHAIN_NONE:
 		default:
 			break;
diff --git a/tools/perf/util/sort.c b/tools/perf/util/sort.c
index 40c9acd41cad3d59cdc32788707567b50ddab159..60ced707bd6b5a88993fb206e098fe1fcee9cd72 100644
--- a/tools/perf/util/sort.c
+++ b/tools/perf/util/sort.c
@@ -5,8 +5,9 @@ char		default_parent_pattern[] = "^sys_|^do_page_fault";
 char		*parent_pattern = default_parent_pattern;
 char		default_sort_order[] = "comm,dso,symbol";
 char		*sort_order = default_sort_order;
-int sort__need_collapse = 0;
-int sort__has_parent = 0;
+int		sort__need_collapse = 0;
+int		sort__has_parent = 0;
+int		sort_by_sym_first;
 
 unsigned int dsos__col_width;
 unsigned int comms__col_width;
@@ -265,6 +266,10 @@ int sort_dimension__add(const char *tok)
 			sort__has_parent = 1;
 		}
 
+		if (list_empty(&hist_entry__sort_list) &&
+		    !strcmp(sd->name, "symbol"))
+			sort_by_sym_first = true;
+
 		list_add_tail(&sd->entry->list, &hist_entry__sort_list);
 		sd->taken = 1;
 
@@ -273,4 +278,3 @@ int sort_dimension__add(const char *tok)
 
 	return -ESRCH;
 }
-
diff --git a/tools/perf/util/sort.h b/tools/perf/util/sort.h
index 13806d782af6b1c823d2886c73ad7acee6305148..24c2b709f0d3921c420c966fe53aad11f94280e3 100644
--- a/tools/perf/util/sort.h
+++ b/tools/perf/util/sort.h
@@ -39,6 +39,7 @@ extern struct sort_entry sort_parent;
 extern unsigned int dsos__col_width;
 extern unsigned int comms__col_width;
 extern unsigned int threads__col_width;
+extern int sort_by_sym_first;
 
 struct hist_entry {
 	struct rb_node		rb_node;