Re: [PATCH 2/2] spi: qup: Request CS GPIO's during probe

["Ivan T. Ivanov" <iivanov@mm-sol.com>]

On Tue, 2015-03-10 at 11:06 +0000, Mark Brown wrote:
> On Tue, Mar 10, 2015 at 10:10:56AM +0200, Ivan T. Ivanov wrote:
> > On Mon, 2015-03-09 at 18:28 +0000, Mark Brown wrote:
> 
> > > > About the API usage, point taken. GPIO requesting part is more important
> > > > in this case. pinctrl core did not request pins and wrong DT configuration
> > > > could lead to surprises without any warnings or errors.
> 
> > > That doesn't answer my concern at all.
> 
> > I am not sure that I am following you.
> 
> > I can not use spi_master::cs_gpios, which is populated by
> > of_spi_register_master(), because spi_register_master()
> > populate SPI devices and they could issue setup method.
> 
> I'm sorry but I can't parse the above.  What does "they could issue
> setup method" mean and why is it a problem?

Client drivers could execute spi_setup() in probe(), so we have
to ensure that CS GPIO's are available before this, no?

> 
> > Requesting GPIO's in core framework is also not a easy
> > option because of arguments here[1].
> 
> We should really fix that though.
> 

I think that pinctrl framework should automatically
request pins belonging to group when state is selected.

Regards,
Ivan