Re: [PATCH] OLPC: ALSA: fix cs5535audio's MIC GPIO to enable input

[Ben Gardner <gardner.ben@gmail.com>]

Hi Andres,

> Previously the MIC GPIO was set to output mode, and when checking the status
> after setting it we were checking OUTPUT_VAL.  This worked, though I'm not
> quite sure why. Instead, if we actually check the READ_BACK value, it
> doesn't work unless the GPIO is in bidirectional mode.  Thus, enable
> input mode as well.

The cs5535-gpio driver was reading output value, not the sensed input value.
When that was fixed, this OLPC code failed because the input wasn't
enabled, so the gpio_get_value() call didn't return anything useful.
At reset the CS5535/6 GPIOs have both input and output disabled.

> Signed-off-by: Andres Salomon <dilinger@collabora.co.uk>
> ---
>  sound/pci/cs5535audio/cs5535audio_olpc.c |    1 +
>  1 files changed, 1 insertions(+), 0 deletions(-)
>
> diff --git a/sound/pci/cs5535audio/cs5535audio_olpc.c b/sound/pci/cs5535audio/cs5535audio_olpc.c
> index 50da49b..f5574f2 100644
> --- a/sound/pci/cs5535audio/cs5535audio_olpc.c
> +++ b/sound/pci/cs5535audio/cs5535audio_olpc.c
> @@ -157,6 +157,7 @@ int __devinit olpc_quirks(struct snd_card *card, struct snd_ac97 *ac97)
>                return -EIO;
>        }
>        gpio_direction_output(OLPC_GPIO_MIC_AC, 0);
> +       gpio_direction_input(OLPC_GPIO_MIC_AC);
>
>        /* drop the original AD1888 HPF control */
>        memset(&elem, 0, sizeof(elem));

This will only work because the cs5535-gpio driver and gpiolib are
both 'broken'.

The problem with gpiolib is that it only allows a GPIO pin to be
either an input or output, but not both.
It has two separate functions to configure the direction
(gpio_direction_input/gpio_direction_output), where one should do (ie,
gpio_set_direction).
>From what I can tell, when direction_input() is called, the GPIO chip
is expected to disable the output and enable the input. If that really
occurred, then the above code would still be broken.

The cs5535-gpio driver doesn't follow that input-or-output convention.
It never disables the direction that wasn't requested.

I'm working on a gpiolib patch that will combine gpio_direction_output
and gpio_direction_input into one function.  That would enable the
above driver to be 'obviously correct'.

Ben