[patch 2.6.28-rc5] regulator: enable/disable refcounting

[patch 2.6.28-rc5] regulator: enable/disable refcounting

[David Brownell]

From: David Brownell <dbrownell@users.sourceforge.net>

Make the <linux/regulator.h> framework treat enable/disable call
pairs like the <linux/clk.h> and <linux/interrupt.h> frameworks do:
they're refcounted, so that different parts of a driver don't need
to put work into coordination that frameworks normally handle.
It's a minor object code shrink.

It also makes the regulator_is_disabled() kerneldoc say what it's
actually returning:  return value is not a refcount, and may report
an error (e.g. I/O error from I2C).

It also fixes some minor regulator_put() goofage:  removing unlocked
access to the enable state.  (But still not making regulator put/get
match the refcounting pattern they invoke.)

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
---
 drivers/regulator/core.c |   71 ++++++++++++++++++++++-----------------------
 1 file changed, 36 insertions(+), 35 deletions(-)

--- a/drivers/regulator/core.c
+++ b/drivers/regulator/core.c
@@ -79,7 +79,7 @@ struct regulator {
 	int uA_load;
 	int min_uV;
 	int max_uV;
-	int enabled; /* client has called enabled */
+	int enabled; /* count of client enables */
 	char *supply_name;
 	struct device_attribute dev_attr;
 	struct regulator_dev *rdev;
@@ -933,16 +933,13 @@ void regulator_put(struct regulator *reg
 	if (regulator == NULL || IS_ERR(regulator))
 		return;
 
-	if (regulator->enabled) {
-		printk(KERN_WARNING "Releasing supply %s while enabled\n",
-		       regulator->supply_name);
-		WARN_ON(regulator->enabled);
-		regulator_disable(regulator);
-	}
-
 	mutex_lock(&regulator_list_mutex);
 	rdev = regulator->rdev;
 
+	if (WARN(regulator->enabled, "Releasing supply %s while enabled\n",
+			       regulator->supply_name))
+		_regulator_disable(rdev);
+
 	/* remove any sysfs entries */
 	if (regulator->dev) {
 		sysfs_remove_link(&rdev->dev.kobj, regulator->supply_name);
@@ -1012,21 +1009,17 @@ static int _regulator_enable(struct regu
  */
 int regulator_enable(struct regulator *regulator)
 {
-	int ret;
-
-	if (regulator->enabled) {
-		printk(KERN_CRIT "Regulator %s already enabled\n",
-		       regulator->supply_name);
-		WARN_ON(regulator->enabled);
-		return 0;
-	}
+	struct regulator_dev *rdev = regulator->rdev;
+	int ret = 0;
 
-	mutex_lock(&regulator->rdev->mutex);
-	regulator->enabled = 1;
-	ret = _regulator_enable(regulator->rdev);
-	if (ret != 0)
-		regulator->enabled = 0;
-	mutex_unlock(&regulator->rdev->mutex);
+	mutex_lock(&rdev->mutex);
+	if (regulator->enabled == 0)
+		ret = _regulator_enable(rdev);
+	else if (regulator->enabled < 0)
+		ret = -EIO;
+	if (ret == 0)
+		regulator->enabled++;
+	mutex_unlock(&rdev->mutex);
 	return ret;
 }
 EXPORT_SYMBOL_GPL(regulator_enable);
@@ -1078,19 +1071,21 @@ static int _regulator_disable(struct reg
  */
 int regulator_disable(struct regulator *regulator)
 {
-	int ret;
-
-	if (!regulator->enabled) {
-		printk(KERN_ERR "%s: not in use by this consumer\n",
-			__func__);
-		return 0;
-	}
+	struct regulator_dev *rdev = regulator->rdev;
+	int ret = 0;
 
-	mutex_lock(&regulator->rdev->mutex);
-	regulator->enabled = 0;
-	regulator->uA_load = 0;
-	ret = _regulator_disable(regulator->rdev);
-	mutex_unlock(&regulator->rdev->mutex);
+	mutex_lock(&rdev->mutex);
+	if (regulator->enabled == 1) {
+		ret = _regulator_disable(rdev);
+		if (ret == 0)
+			regulator->uA_load = 0;
+	} else if (WARN(regulator->enabled <= 0,
+			"unbalanced disables for supply %s\n",
+			regulator->supply_name))
+		ret = -EIO;
+	if (ret == 0)
+		regulator->enabled--;
+	mutex_unlock(&rdev->mutex);
 	return ret;
 }
 EXPORT_SYMBOL_GPL(regulator_disable);
@@ -1166,7 +1161,13 @@ out:
  * regulator_is_enabled - is the regulator output enabled
  * @regulator: regulator source
  *
- * Returns zero for disabled otherwise return number of enable requests.
+ * Returns positive if the regulator driver backing the source/client
+ * has requested that the device be enabled, zero if it hasn't, else a
+ * negative errno code.
+ *
+ * Note that the device backing this regulator handle can have multiple
+ * users, so it might be enabled even if regulator_enable() was never
+ * called for this particular source.
  */
 int regulator_is_enabled(struct regulator *regulator)
 {

Re: [patch 2.6.28-rc5] regulator: enable/disable refcounting

[Mark Brown]

On Sun, Nov 16, 2008 at 11:44:46AM -0800, David Brownell wrote:

> +	mutex_lock(&rdev->mutex);
> +	if (regulator->enabled == 1) {
> +		ret = _regulator_disable(rdev);
> +		if (ret == 0)
> +			regulator->uA_load = 0;

Why clear the load here?  I'd expect to be able to enable and disable
the regulator while preserving the configuration.

Re: [patch 2.6.28-rc5] regulator: enable/disable refcounting

[David Brownell]

On Monday 17 November 2008, Mark Brown wrote:
> On Sun, Nov 16, 2008 at 11:44:46AM -0800, David Brownell wrote:
> 
> > +	mutex_lock(&rdev->mutex);
> > +	if (regulator->enabled == 1) {
> > +		ret = _regulator_disable(rdev);
> > +		if (ret == 0)
> > +			regulator->uA_load = 0;
> 
> Why clear the load here?  I'd expect to be able to enable and disable
> the regulator while preserving the configuration.

That clear was in the original code.  Didn't make much sense
to me either, but I was just caring it forward...

Re: [patch 2.6.28-rc5] regulator: enable/disable refcounting

[Liam Girdwood]

On Mon, 2008-11-17 at 08:21 -0800, David Brownell wrote:
> On Monday 17 November 2008, Mark Brown wrote:
> > On Sun, Nov 16, 2008 at 11:44:46AM -0800, David Brownell wrote:
> > 
> > > +	mutex_lock(&rdev->mutex);
> > > +	if (regulator->enabled == 1) {
> > > +		ret = _regulator_disable(rdev);
> > > +		if (ret == 0)
> > > +			regulator->uA_load = 0;
> > 
> > Why clear the load here?  I'd expect to be able to enable and disable
> > the regulator while preserving the configuration.
> 
> That clear was in the original code.  Didn't make much sense
> to me either, but I was just caring it forward...
> 

Hmm, this looks like a bug - I'll fix in a subsequent patch.

Will have time to look over the rest later tonight.

Thanks

Liam 

Re: [patch 2.6.28-rc5] regulator: enable/disable refcounting

[Liam Girdwood]

On Sun, 2008-11-16 at 11:44 -0800, David Brownell wrote:
> From: David Brownell <dbrownell@users.sourceforge.net>

Both -rc5 patches applied.

Thanks

Liam

Re: [patch 2.6.28-rc5] regulator: enable/disable refcounting

[David Brownell]

On Wednesday 19 November 2008, Liam Girdwood wrote:
> Both -rc5 patches applied.

How about the other two which seemed non-controversial?

  http://marc.info/?l=linux-kernel&m=122645403604873&w=2 (code shrink)
  http://marc.info/?l=linux-kernel&m=122645416305013&w=2 (attribute shrink)

Those were sent against RC4.

- Dave

Re: [patch 2.6.28-rc5] regulator: enable/disable refcounting

[Liam Girdwood]

On Wed, 2008-11-19 at 14:12 -0800, David Brownell wrote:
> On Wednesday 19 November 2008, Liam Girdwood wrote:
> > Both -rc5 patches applied.
> 
> How about the other two which seemed non-controversial?
> 
>   http://marc.info/?l=linux-kernel&m=122645403604873&w=2 (code shrink)
>   http://marc.info/?l=linux-kernel&m=122645416305013&w=2 (attribute shrink)
> 
> Those were sent against RC4.

Both applied.

Thanks

Liam

Re: [patch 2.6.28-rc5] regulator: enable/disable refcounting

[David Brownell]

On Thursday 20 November 2008, Liam Girdwood wrote:
> On Wed, 2008-11-19 at 14:12 -0800, David Brownell wrote:
> > On Wednesday 19 November 2008, Liam Girdwood wrote:
> > > Both -rc5 patches applied.
> > 
> > How about the other two which seemed non-controversial?
> > 
> >   http://marc.info/?l=linux-kernel&m=122645403604873&w=2 (code shrink)
> >   http://marc.info/?l=linux-kernel&m=122645416305013&w=2 (attribute shrink)
> > 
> > Those were sent against RC4.
> 
> Both applied.

Great!  Now, I think I have some comments from Mark to respond to ...
He seems to be saying a new regulator_ops method is needed.

- Dave