Re: [PATCH 2/2] (v2) Mark MCP23S08 as one that will not sleep.

[Alexander Stein <alexander.stein@xxxxxxxxxxxxxxxxxxxxx>]

Am Montag, 13. Mai 2019, 14:00:24 CEST schrieb Joe Burmeister:
> Though it has a 'standby' it doesn't appear to be an issue and
> marking the chip with can_sleep means gpiolib.c won't allow its use
> as a interrupt controller.
> ---
>  drivers/pinctrl/pinctrl-mcp23s08.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/pinctrl/pinctrl-mcp23s08.c b/drivers/pinctrl/pinctrl-mcp23s08.c
> index 3fc63cb5b332..7334d8eb9135 100644
> --- a/drivers/pinctrl/pinctrl-mcp23s08.c
> +++ b/drivers/pinctrl/pinctrl-mcp23s08.c
> @@ -890,7 +890,7 @@ static int mcp23s08_probe_one(struct mcp23s08 *mcp, struct device *dev,
>  		return PTR_ERR(mcp->regmap);
>  
>  	mcp->chip.base = base;
> -	mcp->chip.can_sleep = true;
> +	mcp->chip.can_sleep = false;
>  	mcp->chip.parent = dev;
>  	mcp->chip.owner = THIS_MODULE;
>  
> 

IMHO this is completly wrong, please refer to the documentation of this flag, e.g. https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/include/linux/gpio/driver.h#n217
It essentially means you can't use this GPIOs from atomic context, as SPI or I2C transfers block/sleep, hence the name.
In your case the IRQs are probably not requested as threaded, as stated in the link above.

Best regards,
Alexander