Re: [PATCH] driver: gpio: fix gpio glitch for output-high gpio-hog

On Wed, Apr 6, 2022 at 5:27 AM Tom Rini trini@konsulko.com wrote:

On Wed, Mar 02, 2022 at 06:01:08PM -0800, Tim Harvey wrote:

...

A gpio-hog can be specified as output-low, output-high, or input where output-low means 'de-asserted' and 'output-high' means asserted vs voltage levels.

When a hog is probed gpio_request_tail() calls dm_gpio_set_dir_flags() which ends up setting the GPIO as an output with a driven 'de-asserted' value prior to setting the desired value the hog was configured for. While I'm not sure it makes sense to set the output level while simply 'requesting' a GPIO the result of this is that if the hog is configured for output-high the request call sets it first as output low before gpio_hog_probe() sets it to the configured value causing the gpio to 'glitch' which may be undesired for certain applications.

Fix this by setting the GPIOD_IS_OUT_ACTIVE flag for hogs configured as output-high.

This was tested with the following hogs:

    /* active-high output-low (de-asserted) GPIO should drive 0 */
    gpio1 {
            gpio-hog;
            output-low;
            gpios = <1 GPIO_ACTIVE_HIGH>;
            line-name = "gpio1";
    };

    /* active-high output-high (asserted) GPIO should drive 1 */
  /* before patch this would first drive 0 then 1 */
    gpio2 {
            gpio-hog;
            output-high;
            gpios = <2 GPIO_ACTIVE_HIGH>;
            line-name = "gpio2";
    };

    /* active-low output-low (de-asserted) GPIO should drive 1 */
    gpio3 {
            gpio-hog;
            output-low;
            gpios = <3 GPIO_ACTIVE_LOW>;
            line-name = "gpio3#";
    };

    /* active-low output-high (asserted) GPIO should drive 0 */
  /* before patch this would first drive 0 then 1 */
    gpio4 {
            gpio-hog;
            output-high;
            gpios = <4 GPIO_ACTIVE_LOW>;
            line-name = "gpio4#";
    };

Cc: Sean Anderson sean.anderson@seco.com Signed-off-by: Tim Harvey tharvey@gateworks.com

This breaks the ut_dm_dm_test_gpio test which I guess needs to be updated.

Tom,

That's kind of funny I suppose because the behavior is clearly wrong.

Is fixing the test something that needs to be done for the patch to be applied, or needs to be done by me? I'm not familiar with the test framework or how to execute it.

Tim