Skip to contentSkip to conversation

regulator: check for devicetree early.

PATCH In review NeilBrown 2 participants 3 messages Nov 1, 2013 Voltage and current regulator framework

lore
Share & tools
lore

NeilBrown

author
#0 lore raw

PATCH 1/1 regulator: check for devicetree early.

The behaviour of regulator core with respect to device-tree and CONFIG_REGULATOR_DUMMY is inconsistent.

If there is no device-tree, then a board file can call regulator_has_full_constraints() and this will negate the effect of CONFIG_REGULATOR_DUMMY.

If a device-tree is given, regulator_has_full_constraints() is called automatically, but too late to effectively negate the effect of CONFIG_REGULATOR_DUMMY.

So it makes sense for regulator/core to check for devicetree earlier and call regulator_has_full_constraints() before any call to regulator_get() can trip up on CONFIG_REGULATOR_DUMMY.

drivers/regulator/core.c +11 -6
index a01b8b3b70ca..8cd457ae66bc 100644
@@ -3748,13 +3748,8 @@ static int __init regulator_init(void)
 /* init early to allow our consumers to complete system booting */
 core_initcall(regulator_init);
 
-static int __init regulator_init_complete(void)
+static int __init regulator_init_early(void)
 {
-	struct regulator_dev *rdev;
-	struct regulator_ops *ops;
-	struct regulation_constraints *c;
-	int enabled, ret;
-
 	/*
 	 * Since DT doesn't provide an idiomatic mechanism for
 	 * enabling full constraints and since it's much more natural
@@ -3763,6 +3758,16 @@ static int __init regulator_init_complete(void)
 	 */
 	if (of_have_populated_dt())
 		has_full_constraints = true;
+	return 0;
+}
+early_initcall(regulator_init_early);
+
+static int __init regulator_init_complete(void)
+{
+	struct regulator_dev *rdev;
+	struct regulator_ops *ops;
+	struct regulation_constraints *c;
+	int enabled, ret;
 
 	mutex_lock(&regulator_list_mutex);
 
maintainer
#1 lore raw
Quoted 2 lines · NeilBrown

The behaviour of regulator core with respect to device-tree and CONFIG_REGULATOR_DUMMY is inconsistent.

Wei Ni reported this issue and is working on a patch which I'm expecting to do this as part of device tree initialisation rather than by faffing around with initcall ordering; this will be simpler and more robust since there won't be any possibility of races being introduced.

author
#2 lore raw
Quoted 9 lines · On Thu, 31 Oct 2013 23:42:40 -0700 Mark Brown <[email protected]> wrote:
Quoted 2 lines · NeilBrown

The behaviour of regulator core with respect to device-tree and CONFIG_REGULATOR_DUMMY is inconsistent.

Wei Ni reported this issue and is working on a patch which I'm expecting to do this as part of device tree initialisation rather than by faffing around with initcall ordering; this will be simpler and more robust since there won't be any possibility of races being introduced.

OK, thanks. Sounds like a better approach.

NeilBrown