[PATCH] gpio: twl4030: fix regression for twl gpio outputs for led pins

[PATCH] gpio: twl4030: fix regression for twl gpio outputs for led pins

[Nishanth Menon]

Commit 0b2aa8b (gpio: twl4030: Fix regression for twl gpio output)
introduces yet another regression by ignoring the fact that we use twl
gpio even for LED. For example, leda (offset 18) is used for USB port2
power on beagleboard-xm.

The original logic was supposed to setup the TWL4030 GPIO direction
if the offset was < TWL4030_GPIO_MAX, and anything greater used
twl4030_led_set_value as part of twl_set to setup the LED pin. However,
with the commit mentioned above, LEDa pin request will result in fail
as the GPIO number for LEDa is equal to TWL4030_GPIO_MAX and USB fails
to function on beagleboard-xm. This is reported in the log as:
hsusb2_vbus: Failed to request enable GPIO510: -22

To fix this, we should handle the fail case of
twl4030_set_gpio_direction independent of the offset < TWL4030_GPIO_MAX
check.

Signed-off-by: Nishanth Menon <nm@ti.com>
---
Applies on v3.13-rc3

Tested on beagleboard-xm which uses ethernet over usb:
before the patch (USB regression): http://pastebin.mozilla.org/3769443
After this patch: http://pastebin.mozilla.org/3769454

 drivers/gpio/gpio-twl4030.c |    9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/gpio/gpio-twl4030.c b/drivers/gpio/gpio-twl4030.c
index b97d6a6..9242276 100644
--- a/drivers/gpio/gpio-twl4030.c
+++ b/drivers/gpio/gpio-twl4030.c
@@ -354,11 +354,16 @@ static void twl_set(struct gpio_chip *chip, unsigned offset, int value)
 static int twl_direction_out(struct gpio_chip *chip, unsigned offset, int value)
 {
 	struct gpio_twl4030_priv *priv = to_gpio_twl4030(chip);
-	int ret = -EINVAL;
+	int ret = 0;
 
 	mutex_lock(&priv->mutex);
-	if (offset < TWL4030_GPIO_MAX)
+	if (offset < TWL4030_GPIO_MAX) {
 		ret = twl4030_set_gpio_direction(offset, 0);
+		if (ret) {
+			mutex_unlock(&priv->mutex);
+			return ret;
+		}
+	}
 
 	priv->direction |= BIT(offset);
 	mutex_unlock(&priv->mutex);
-- 
1.7.9.5

Re: [PATCH] gpio: twl4030: fix regression for twl gpio outputs for led pins

[Tony Lindgren]

* Nishanth Menon <nm@ti.com> [131213 08:48]:
> Commit 0b2aa8b (gpio: twl4030: Fix regression for twl gpio output)
> introduces yet another regression by ignoring the fact that we use twl
> gpio even for LED. For example, leda (offset 18) is used for USB port2
> power on beagleboard-xm.
> 
> The original logic was supposed to setup the TWL4030 GPIO direction
> if the offset was < TWL4030_GPIO_MAX, and anything greater used
> twl4030_led_set_value as part of twl_set to setup the LED pin. However,
> with the commit mentioned above, LEDa pin request will result in fail
> as the GPIO number for LEDa is equal to TWL4030_GPIO_MAX and USB fails
> to function on beagleboard-xm. This is reported in the log as:
> hsusb2_vbus: Failed to request enable GPIO510: -22
> 
> To fix this, we should handle the fail case of
> twl4030_set_gpio_direction independent of the offset < TWL4030_GPIO_MAX
> check.

Heh again :) Linus has already queued a similar fix, see
"[PATCH v2 1/1] gpio: twl4030: Fix regression for twl gpio LED output".

Regards,

Tony



 
> Signed-off-by: Nishanth Menon <nm@ti.com>
> ---
> Applies on v3.13-rc3
> 
> Tested on beagleboard-xm which uses ethernet over usb:
> before the patch (USB regression): http://pastebin.mozilla.org/3769443
> After this patch: http://pastebin.mozilla.org/3769454
> 
>  drivers/gpio/gpio-twl4030.c |    9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpio/gpio-twl4030.c b/drivers/gpio/gpio-twl4030.c
> index b97d6a6..9242276 100644
> --- a/drivers/gpio/gpio-twl4030.c
> +++ b/drivers/gpio/gpio-twl4030.c
> @@ -354,11 +354,16 @@ static void twl_set(struct gpio_chip *chip, unsigned offset, int value)
>  static int twl_direction_out(struct gpio_chip *chip, unsigned offset, int value)
>  {
>  	struct gpio_twl4030_priv *priv = to_gpio_twl4030(chip);
> -	int ret = -EINVAL;
> +	int ret = 0;
>  
>  	mutex_lock(&priv->mutex);
> -	if (offset < TWL4030_GPIO_MAX)
> +	if (offset < TWL4030_GPIO_MAX) {
>  		ret = twl4030_set_gpio_direction(offset, 0);
> +		if (ret) {
> +			mutex_unlock(&priv->mutex);
> +			return ret;
> +		}
> +	}
>  
>  	priv->direction |= BIT(offset);
>  	mutex_unlock(&priv->mutex);
> -- 
> 1.7.9.5
> 

Re: [PATCH] gpio: twl4030: fix regression for twl gpio outputs for led pins

[Nishanth Menon]

On 12/13/2013 11:28 AM, Tony Lindgren wrote:
> * Nishanth Menon <nm@ti.com> [131213 08:48]:
>> Commit 0b2aa8b (gpio: twl4030: Fix regression for twl gpio output)
>> introduces yet another regression by ignoring the fact that we use twl
>> gpio even for LED. For example, leda (offset 18) is used for USB port2
>> power on beagleboard-xm.
>>
>> The original logic was supposed to setup the TWL4030 GPIO direction
>> if the offset was < TWL4030_GPIO_MAX, and anything greater used
>> twl4030_led_set_value as part of twl_set to setup the LED pin. However,
>> with the commit mentioned above, LEDa pin request will result in fail
>> as the GPIO number for LEDa is equal to TWL4030_GPIO_MAX and USB fails
>> to function on beagleboard-xm. This is reported in the log as:
>> hsusb2_vbus: Failed to request enable GPIO510: -22
>>
>> To fix this, we should handle the fail case of
>> twl4030_set_gpio_direction independent of the offset < TWL4030_GPIO_MAX
>> check.
> 
> Heh again :) Linus has already queued a similar fix, see
> "[PATCH v2 1/1] gpio: twl4030: Fix regression for twl gpio LED output".
> 

Alrite, I admit it, I need a vacation ;).. Sigh..
http://marc.info/?l=linux-gpio&w=2&r=1&s=twl4030&q=t is not good
enough :(..


-- 
Regards,
Nishanth Menon