On Fri, Dec 11, 2020 at 12:34:52AM +0800, Tzung-Bi Shih wrote:
On Thu, Dec 10, 2020 at 11:40 PM Mark Brown broonie@kernel.org wrote:
Now this is in DAPM it's not in atomic context so it'd be more friendly to use gpiod_set_value_cansleep() so that it'll work even if the GPIO isn't atomic safe.
Strange, I thought I should get a warning message because of the following statement in gpiod_set_value (but it didn't): WARN_ON(desc->gdev->chip->can_sleep);
Also I dumped the value from gpiod_cansleep(rt1015p->sdb), it returned 0.
Is it safe to call gpiod_set_value_cansleep() even if the GPIO controller doesn't support it? (I guess yes, _cansleep could be just an advice)
Yes, absolutely - it's the other way around. _cansleep() is the caller saying that they support using a GPIO that might sleep when updating (eg, one connected over I2C), it works with both GPIOs that might sleep and GPIOs that are atomic safe. Without _cansleep() only atomic safe GPIO controllers can be used.