Re: [patch] gpiolib: fix oops in gpio_get_value_cansleep()

[David Brownell <david-b@pacbell.net>]

On Thursday 16 October 2008, Anton Vorontsov wrote:
> > -     return chip->get(chip, gpio - chip->base);
> > +     return chip->get ? chip->get(chip, gpio - chip->base) : 0;
> 
> Why don't we check the .set in the gpio_set_value? Because
> we must always call gpio_direction_output()?

Yes.  The output driver needs to be explicitly enabled,
to avoid misbehaving electic circuitry.


> It is not exactly the 
> same we work with the input direction.. is this documented anywhere?

In Documentation/gpio.txt since the very first versions.

See the section on "Spinlock-Safe GPIO access", where
special cases for reading the value of output GPIOs are
desribed.  Other aspects are mentioned in other spots.

For example, open drain signals -- as used with I2C and
other protocols -- only drive the "low" signal level,
and if code wants a "high" level it's got to verify that
nobody else is driving that shared line to low.  By
reading it back, even though it's configured as output.

- Dave