[PATCH 1/2] mfd: madera: Improve handling of regulator unbinding

[PATCH 1/2] mfd: madera: Improve handling of regulator unbinding

[Charles Keepax]

The current unbinding process for Madera has some issues. The trouble
is runtime PM is disabled as the first step of the process, but
some of the drivers release IRQs causing regmap IRQ to issue a
runtime get which fails. To allow runtime PM to remain enabled during
mfd_remove_devices, the DCVDD regulator must remain available. In
the case of external DCVDD's this is simple, the regulator can simply
be disabled/put after the call to mfd_remove_devices. However, in
the case of an internally supplied DCVDD the regulator needs to be
released after all the MFD children, except for the regulator child
itself, have been removed. This is achieved by having the regulator
driver itself do the disable/put, as it is the last driver removed from
the MFD.

Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
---
 drivers/mfd/madera-core.c | 20 ++++++++++++--------
 1 file changed, 12 insertions(+), 8 deletions(-)

diff --git a/drivers/mfd/madera-core.c b/drivers/mfd/madera-core.c
index a8cfadc1fc01e..5170b55836fc1 100644
--- a/drivers/mfd/madera-core.c
+++ b/drivers/mfd/madera-core.c
@@ -730,18 +730,22 @@ int madera_dev_exit(struct madera *madera)
 	/* Prevent any IRQs being serviced while we clean up */
 	disable_irq(madera->irq);
 
-	/*
-	 * DCVDD could be supplied by a child node, we must disable it before
-	 * removing the children, and prevent PM runtime from turning it back on
-	 */
+	pm_runtime_get_sync(madera->dev);
+
+	mfd_remove_devices(madera->dev);
+
 	pm_runtime_disable(madera->dev);
 
-	clk_disable_unprepare(madera->mclk[MADERA_MCLK2].clk);
+	if (!madera->internal_dcvdd) {
+		regulator_disable(madera->dcvdd);
+		regulator_put(madera->dcvdd);
+	}
 
-	regulator_disable(madera->dcvdd);
-	regulator_put(madera->dcvdd);
+	pm_runtime_set_suspended(madera->dev);
+	pm_runtime_put_noidle(madera->dev);
+
+	clk_disable_unprepare(madera->mclk[MADERA_MCLK2].clk);
 
-	mfd_remove_devices(madera->dev);
 	madera_enable_hard_reset(madera);
 
 	regulator_bulk_disable(madera->num_core_supplies,
-- 
2.11.0

[PATCH 2/2] regulator: arizona-ldo1: Improve handling of regulator unbinding

[Charles Keepax]

The current unbinding process for Madera has some issues. The trouble
is runtime PM is disabled as the first step of the process, but
some of the drivers release IRQs causing regmap IRQ to issue a
runtime get which fails. To allow runtime PM to remain enabled during
mfd_remove_devices, the DCVDD regulator must remain available. In
the case of external DCVDD's this is simple, the regulator can simply
be disabled/put after the call to mfd_remove_devices. However, in
the case of an internally supplied DCVDD the regulator needs to be
released after all the MFD children, except for the regulator child
itself, have been removed. This is achieved by having the regulator
driver itself do the disable/put, as it is the last driver removed from
the MFD.

Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
---
 drivers/regulator/arizona-ldo1.c | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/drivers/regulator/arizona-ldo1.c b/drivers/regulator/arizona-ldo1.c
index 1a3d7b720f5e0..83dd37dbfe07b 100644
--- a/drivers/regulator/arizona-ldo1.c
+++ b/drivers/regulator/arizona-ldo1.c
@@ -375,6 +375,18 @@ static int madera_ldo1_probe(struct platform_device *pdev)
 	return 0;
 }
 
+static int madera_ldo1_remove(struct platform_device *pdev)
+{
+	struct madera *madera = dev_get_drvdata(pdev->dev.parent);
+
+	if (madera->internal_dcvdd) {
+		regulator_disable(madera->dcvdd);
+		regulator_put(madera->dcvdd);
+	}
+
+	return arizona_ldo1_remove(pdev);
+}
+
 static struct platform_driver arizona_ldo1_driver = {
 	.probe = arizona_ldo1_probe,
 	.remove = arizona_ldo1_remove,
@@ -385,7 +397,7 @@ static struct platform_driver arizona_ldo1_driver = {
 
 static struct platform_driver madera_ldo1_driver = {
 	.probe = madera_ldo1_probe,
-	.remove = arizona_ldo1_remove,
+	.remove = madera_ldo1_remove,
 	.driver		= {
 		.name	= "madera-ldo1",
 	},
-- 
2.11.0

Re: [PATCH 1/2] mfd: madera: Improve handling of regulator unbinding

[Lee Jones]

On Mon, 09 Dec 2019, Charles Keepax wrote:

> The current unbinding process for Madera has some issues. The trouble
> is runtime PM is disabled as the first step of the process, but
> some of the drivers release IRQs causing regmap IRQ to issue a
> runtime get which fails. To allow runtime PM to remain enabled during
> mfd_remove_devices, the DCVDD regulator must remain available. In
> the case of external DCVDD's this is simple, the regulator can simply
> be disabled/put after the call to mfd_remove_devices. However, in
> the case of an internally supplied DCVDD the regulator needs to be
> released after all the MFD children, except for the regulator child
> itself, have been removed. This is achieved by having the regulator
> driver itself do the disable/put, as it is the last driver removed from
> the MFD.
> 
> Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
> ---
>  drivers/mfd/madera-core.c | 20 ++++++++++++--------
>  1 file changed, 12 insertions(+), 8 deletions(-)

I'm okay with it if Mark is:

For my own reference:
  Acked-for-MFD-by: Lee Jones <lee.jones@linaro.org>

-- 
Lee Jones [李琼斯]
Linaro Services Technical Lead
Linaro.org │ Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog