regmap-mmio and paged registers

[Alexandre Belloni]

On 09/12/2015 at 22:10:36 +0000, Mark Brown wrote :
> On Wed, Dec 09, 2015 at 10:41:50PM +0100, Alexandre Belloni wrote:
> 
> > 	/* It is possible to have selector register inside data window.
> > 	   In that case, selector register is located on every page and
> > 	   it needs no page switching, when accessed alone. */
> > 	if (val_num > 1 ||
> > 	    range->window_start + win_offset != range->selector_reg) {
> 
> > So it ends up never writing the page number in the register.
> 
> > One possible solution would be to implement our own .read and .write to
> > handle that paging but maybe you can think of something else.
> 
> So to be explicit about the actual issue the problem here is that you
> have data bits in the same register as your selector and indeed the
> register map you're trying to page is a single register.  That's pretty
> pathological.
> 
> Honestly looking at the register I'm not even convinced it is a paged
> regmap, it looks more like a mailbox for communication with a
> coprocessor than anything else (the fact that the selector is named PID,
> the fact that the bitfields include CMD...).  Are you sure that it's a
> good idea to represent this as a regmap at all?

Yeah, I was probably not clear enough but the regmap actually covers the
whole PMC range and include that particular register. We have to do that
because this range is used by multiple drivers.

This register, PMC_PCR is also used by multiple drivers so it is
important to properly lock between the page selection and the following
read or write.

-- 
Alexandre Belloni, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com