On Wed, Aug 20, 2014 at 04:29:44PM +0100, Catalin Marinas wrote:

On Wed, Aug 20, 2014 at 04:00:28PM +0100, Grant Likely wrote:

quoted

On Thu, 24 Jul 2014 21:00:12 +0800, Hanjun Guo [off-list ref] wrote:

quoted

@@ -345,13 +378,19 @@ int __init psci_init(void)
 	const struct of_device_id *matched_np;
 	psci_initcall_t init_fn;
 
-	np = of_find_matching_node_and_match(NULL, psci_of_match, &matched_np);
+	if (acpi_disabled) {
+		np = of_find_matching_node_and_match(NULL,
+				psci_of_match, &matched_np);
 
-	if (!np)
-		return -ENODEV;
+		if (!np)
+			return -ENODEV;
+
+		init_fn = (psci_initcall_t)matched_np->data;
+
+		return init_fn(np);
+	}
 
-	init_fn = (psci_initcall_t)matched_np->data;
-	return init_fn(np);
+	return psci_0_2_init_acpi();
 }

So, while the functionality is sound, this is kind of a stinky diff. The
ACPI path should not push the OF code into a sub branch. Make it look
like this instead (with the advantage that the diff will be simpler):

+	if (IS_ENABLED(CONFIG_ACPI) && !acpi_disabled)
+		return psci_0_2_init_acpi();

Minor point: we should define acpi_disabled to 1 when !CONFIG_ACPI and
avoid additional IS_ENABLED(CONFIG_ACPI) checks (unless already don this
way).

This is the default behaviour already.

Graeme