[PATCH v3 2/4] clk: add support for clock reparent on set_rate

On 20/05/13 05:15, Stephen Boyd wrote:

Noticed another minor thing. I'm working on the set rate and
parent at the same time op on top of this patch series. Hopefully
have something soon.

On 05/15, James Hogan wrote:

quoted

@@ -1181,6 +1242,11 @@ static void clk_change_rate(struct clk *clk)
 
 	old_rate = clk->rate;
 
+	/* set parent */
+	if (clk->new_parent && clk->new_parent != clk->parent)
+		__clk_set_parent_no_recalc(clk, clk->new_parent,
+					   clk->new_parent_index);
+

This check here for new_parent != clk->parent ...

quoted

@@ -1451,6 +1501,27 @@ static int __clk_set_parent(struct clk *clk, struct clk *parent, u8 p_index)
 	return 0;
 }
 
+static int __clk_set_parent_no_recalc(struct clk *clk, struct clk *parent,
+				      u8 p_index)
+{
+	int ret = 0;
+
+	if (clk->parent == parent)
+		goto out;

causes this to never be true, so we can probably drop it.

quoted

+
+	/* only re-parent if the clock is not in use */
+	ret = __clk_set_parent(clk, parent, p_index);
+	if (ret)
+		goto out;
+
+	/* reparent, but don't propagate rate recalculation downstream */
+	clk_reparent(clk, parent);
+	clk_debug_reparent(clk, parent);
+
+out:
+	return ret;
+}

Also, we never check the return code so it's sort of useless.

Hi Stephen,

Well spotted, and as you pointed it privately it can use
__clk_set_parent() instead of __clk_set_parent_no_recalc() now :)

Thanks
James