[PATCH] net: phy: allow for reset line to be tied to a sleepy GPIO controller

[PATCH] net: phy: allow for reset line to be tied to a sleepy GPIO controller

[Andrea Merello]

mdio_device_reset() makes use of the atomic-pretending API flavor for
handling the PHY reset GPIO line.

I found no hint that mdio_device_reset() is called from atomic context
and indeed it uses usleep_range() since long time, so I would assume that
it is OK to sleep there.

This patch switch to gpiod_set_value_cansleep() in mdio_device_reset().
This is relevant if e.g. the PHY reset line is tied to a I2C GPIO
controller.

This has been tested on a ZynqMP board running an upstream 4.19 kernel and
then hand-ported on current kernel tree.

Signed-off-by: Andrea Merello <andrea.merello@gmail.com>
---
 drivers/net/phy/mdio_device.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/phy/mdio_device.c b/drivers/net/phy/mdio_device.c
index e282600bd83e..c1d345c3cab3 100644
--- a/drivers/net/phy/mdio_device.c
+++ b/drivers/net/phy/mdio_device.c
@@ -121,7 +121,7 @@ void mdio_device_reset(struct mdio_device *mdiodev, int value)
 		return;
 
 	if (mdiodev->reset_gpio)
-		gpiod_set_value(mdiodev->reset_gpio, value);
+		gpiod_set_value_cansleep(mdiodev->reset_gpio, value);
 
 	if (mdiodev->reset_ctrl) {
 		if (value)
-- 
2.17.1

Re: [PATCH] net: phy: allow for reset line to be tied to a sleepy GPIO controller

[Andrew Lunn]

On Fri, Oct 04, 2019 at 03:53:32PM +0200, Andrea Merello wrote:
> mdio_device_reset() makes use of the atomic-pretending API flavor for
> handling the PHY reset GPIO line.
> 
> I found no hint that mdio_device_reset() is called from atomic context
> and indeed it uses usleep_range() since long time, so I would assume that
> it is OK to sleep there.
> 
> This patch switch to gpiod_set_value_cansleep() in mdio_device_reset().
> This is relevant if e.g. the PHY reset line is tied to a I2C GPIO
> controller.
> 
> This has been tested on a ZynqMP board running an upstream 4.19 kernel and
> then hand-ported on current kernel tree.
> 
> Signed-off-by: Andrea Merello <andrea.merello@gmail.com>

Yes, all the other bits of code using GPIOs are using the _cansleep
versions.

Reviewed-by: Andrew Lunn <andrew@lunn.ch>

    Andrew

Re: [PATCH] net: phy: allow for reset line to be tied to a sleepy GPIO controller

[David Miller]

From: Andrea Merello <andrea.merello@gmail.com>
Date: Fri,  4 Oct 2019 15:53:32 +0200

> mdio_device_reset() makes use of the atomic-pretending API flavor for
> handling the PHY reset GPIO line.
> 
> I found no hint that mdio_device_reset() is called from atomic context
> and indeed it uses usleep_range() since long time, so I would assume that
> it is OK to sleep there.
> 
> This patch switch to gpiod_set_value_cansleep() in mdio_device_reset().
> This is relevant if e.g. the PHY reset line is tied to a I2C GPIO
> controller.
> 
> This has been tested on a ZynqMP board running an upstream 4.19 kernel and
> then hand-ported on current kernel tree.
> 
> Signed-off-by: Andrea Merello <andrea.merello@gmail.com>

Applied.