backlight: pwm_bl: re-add driver internal enabled tracking

Message ID 20181109094857.27079-1-heiko@sntech.de
State New
Headers show
Series
  • backlight: pwm_bl: re-add driver internal enabled tracking
Related show

Commit Message

Heiko Stuebner Nov. 9, 2018, 9:48 a.m.
Commit e6bcca0890b9 ("backlight: pwm_bl: Switch to using "atomic" PWM API")
removed the driver internal enabled tracking in favor of simply checking
the pwm state.

This can lead to issues as all of gpio-, regulator- and pwm-state are used
to determine the initial state and the bootloader or kernel can leave them
in an inconsistent state at boot.

In my case on rk3399-kevin, the pwm backlight is build as module and the
kernel disables the supply regulator as unused while keeping the pwm running
thus pwm_bl calling pwm_backlight_power_off() during probe and creating an
unmatched regulator-disable call, as it never got enabled from the pwm-bl
before.

To prevent these consistency issues, reintroduce the driver-internal
tracking of the enabled state.

Fixes: e6bcca0890b9 ("backlight: pwm_bl: Switch to using "atomic" PWM API")
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
---
 drivers/video/backlight/pwm_bl.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

Comments

Enric Balletbo i Serra Nov. 9, 2018, 10:15 a.m. | #1
Hi Heiko,

Many thanks to catch and dig into this, I did more tests but I was not able to
reproduce the issue with my nfs environment. Anyway the patch looks good to me
and fixes the issue for you, so

On 9/11/18 10:48, Heiko Stuebner wrote:
> Commit e6bcca0890b9 ("backlight: pwm_bl: Switch to using "atomic" PWM API")
> removed the driver internal enabled tracking in favor of simply checking
> the pwm state.
> 
> This can lead to issues as all of gpio-, regulator- and pwm-state are used
> to determine the initial state and the bootloader or kernel can leave them
> in an inconsistent state at boot.
> 
> In my case on rk3399-kevin, the pwm backlight is build as module and the
> kernel disables the supply regulator as unused while keeping the pwm running
> thus pwm_bl calling pwm_backlight_power_off() during probe and creating an
> unmatched regulator-disable call, as it never got enabled from the pwm-bl
> before.
> 
> To prevent these consistency issues, reintroduce the driver-internal
> tracking of the enabled state.
> 
> Fixes: e6bcca0890b9 ("backlight: pwm_bl: Switch to using "atomic" PWM API")
> Signed-off-by: Heiko Stuebner <heiko@sntech.de>
> ---
>  drivers/video/backlight/pwm_bl.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/video/backlight/pwm_bl.c b/drivers/video/backlight/pwm_bl.c
> index 678b27063198..bcd08b41765d 100644
> --- a/drivers/video/backlight/pwm_bl.c
> +++ b/drivers/video/backlight/pwm_bl.c
> @@ -30,6 +30,7 @@ struct pwm_bl_data {
>  	struct device		*dev;
>  	unsigned int		lth_brightness;
>  	unsigned int		*levels;
> +	bool			enabled;
>  	struct regulator	*power_supply;
>  	struct gpio_desc	*enable_gpio;
>  	unsigned int		scale;
> @@ -50,7 +51,7 @@ static void pwm_backlight_power_on(struct pwm_bl_data *pb)
>  	int err;
>  
>  	pwm_get_state(pb->pwm, &state);
> -	if (state.enabled)
> +	if (pb->enabled)
>  		return;
>  
>  	err = regulator_enable(pb->power_supply);
> @@ -65,6 +66,8 @@ static void pwm_backlight_power_on(struct pwm_bl_data *pb)
>  
>  	if (pb->enable_gpio)
>  		gpiod_set_value_cansleep(pb->enable_gpio, 1);
> +
> +	pb->enabled = true;
>  }
>  
>  static void pwm_backlight_power_off(struct pwm_bl_data *pb)
> @@ -72,7 +75,7 @@ static void pwm_backlight_power_off(struct pwm_bl_data *pb)
>  	struct pwm_state state;
>  
>  	pwm_get_state(pb->pwm, &state);
> -	if (!state.enabled)
> +	if (!pb->enabled)
>  		return;
>  
>  	if (pb->enable_gpio)
> @@ -86,6 +89,7 @@ static void pwm_backlight_power_off(struct pwm_bl_data *pb)
>  	pwm_apply_state(pb->pwm, &state);
>  
>  	regulator_disable(pb->power_supply);
> +	pb->enabled = false;
>  }
>  
>  static int compute_duty_cycle(struct pwm_bl_data *pb, int brightness)
> @@ -483,6 +487,7 @@ static int pwm_backlight_probe(struct platform_device *pdev)
>  	pb->check_fb = data->check_fb;
>  	pb->exit = data->exit;
>  	pb->dev = &pdev->dev;
> +	pb->enabled = false;
>  	pb->post_pwm_on_delay = data->post_pwm_on_delay;
>  	pb->pwm_off_delay = data->pwm_off_delay;
>  
> 

Reviewed-by: Enric Balletbo i Serra <enric.balletbo@collabora.com>
Uwe Kleine-König Nov. 14, 2018, 9:11 a.m. | #2
Hello Heiko,

On Fri, Nov 09, 2018 at 10:48:57AM +0100, Heiko Stuebner wrote:
> Commit e6bcca0890b9 ("backlight: pwm_bl: Switch to using "atomic" PWM API")
> removed the driver internal enabled tracking in favor of simply checking
> the pwm state.
> 
> This can lead to issues as all of gpio-, regulator- and pwm-state are used
> to determine the initial state and the bootloader or kernel can leave them
> in an inconsistent state at boot.
> 
> In my case on rk3399-kevin, the pwm backlight is build as module and the
> kernel disables the supply regulator as unused while keeping the pwm running
> thus pwm_bl calling pwm_backlight_power_off() during probe and creating an
> unmatched regulator-disable call, as it never got enabled from the pwm-bl
> before.
> 
> To prevent these consistency issues, reintroduce the driver-internal
> tracking of the enabled state.
> 
> Fixes: e6bcca0890b9 ("backlight: pwm_bl: Switch to using "atomic" PWM API")
> Signed-off-by: Heiko Stuebner <heiko@sntech.de>

This looks right,

Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>

Best regards
Uwe
Thierry Reding Nov. 14, 2018, 12:35 p.m. | #3
On Fri, Nov 09, 2018 at 10:48:57AM +0100, Heiko Stuebner wrote:
> Commit e6bcca0890b9 ("backlight: pwm_bl: Switch to using "atomic" PWM API")
> removed the driver internal enabled tracking in favor of simply checking
> the pwm state.
> 
> This can lead to issues as all of gpio-, regulator- and pwm-state are used
> to determine the initial state and the bootloader or kernel can leave them
> in an inconsistent state at boot.
> 
> In my case on rk3399-kevin, the pwm backlight is build as module and the
> kernel disables the supply regulator as unused while keeping the pwm running
> thus pwm_bl calling pwm_backlight_power_off() during probe and creating an
> unmatched regulator-disable call, as it never got enabled from the pwm-bl
> before.
> 
> To prevent these consistency issues, reintroduce the driver-internal
> tracking of the enabled state.
> 
> Fixes: e6bcca0890b9 ("backlight: pwm_bl: Switch to using "atomic" PWM API")
> Signed-off-by: Heiko Stuebner <heiko@sntech.de>
> ---
>  drivers/video/backlight/pwm_bl.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)

Acked-by: Thierry Reding <thierry.reding@gmail.com>

Patch

diff --git a/drivers/video/backlight/pwm_bl.c b/drivers/video/backlight/pwm_bl.c
index 678b27063198..bcd08b41765d 100644
--- a/drivers/video/backlight/pwm_bl.c
+++ b/drivers/video/backlight/pwm_bl.c
@@ -30,6 +30,7 @@  struct pwm_bl_data {
 	struct device		*dev;
 	unsigned int		lth_brightness;
 	unsigned int		*levels;
+	bool			enabled;
 	struct regulator	*power_supply;
 	struct gpio_desc	*enable_gpio;
 	unsigned int		scale;
@@ -50,7 +51,7 @@  static void pwm_backlight_power_on(struct pwm_bl_data *pb)
 	int err;
 
 	pwm_get_state(pb->pwm, &state);
-	if (state.enabled)
+	if (pb->enabled)
 		return;
 
 	err = regulator_enable(pb->power_supply);
@@ -65,6 +66,8 @@  static void pwm_backlight_power_on(struct pwm_bl_data *pb)
 
 	if (pb->enable_gpio)
 		gpiod_set_value_cansleep(pb->enable_gpio, 1);
+
+	pb->enabled = true;
 }
 
 static void pwm_backlight_power_off(struct pwm_bl_data *pb)
@@ -72,7 +75,7 @@  static void pwm_backlight_power_off(struct pwm_bl_data *pb)
 	struct pwm_state state;
 
 	pwm_get_state(pb->pwm, &state);
-	if (!state.enabled)
+	if (!pb->enabled)
 		return;
 
 	if (pb->enable_gpio)
@@ -86,6 +89,7 @@  static void pwm_backlight_power_off(struct pwm_bl_data *pb)
 	pwm_apply_state(pb->pwm, &state);
 
 	regulator_disable(pb->power_supply);
+	pb->enabled = false;
 }
 
 static int compute_duty_cycle(struct pwm_bl_data *pb, int brightness)
@@ -483,6 +487,7 @@  static int pwm_backlight_probe(struct platform_device *pdev)
 	pb->check_fb = data->check_fb;
 	pb->exit = data->exit;
 	pb->dev = &pdev->dev;
+	pb->enabled = false;
 	pb->post_pwm_on_delay = data->post_pwm_on_delay;
 	pb->pwm_off_delay = data->pwm_off_delay;