On Wednesday 22 October 2014 11:51:40 Mika Westerberg wrote:
> On Wed, Oct 22, 2014 at 10:33:32AM +0200, Arnd Bergmann wrote:
> > On Wednesday 22 October 2014 11:10:44 Mika Westerberg wrote:
> > >
> > > It expects that GPIOs returned from _CRS are in specific order. Since we
> > > can't change these existing ACPI tables, we must support them somehow.
> > >
> > > This patch series handles it so that:
> > >
> > > 1) If we can't find given property (e.g "reset-gpios" or
> > > "shutdown-gpios") the index above will refer directly to the GPIO
> > > resource returned from _CRS.
> > >
> > > 2) If the property is found we ignore index and take it from the
> > > property instead.
> > >
> > > This has the drawback that we cannot support this:
> > >
> > > Package () { "reset-gpios", Package () { ^GPIO, 0, 0, 0, ^GPIO, 1, 0, 0}}
> > > ^^^^^^^^^^^^^^
> > > So the second entry in the above is not accessible using
> > > gpiod_get_index() and the reason is that we want to support the existing
> > > and new ACPI tables where _DSD is not being used.
> >
> > So this is not using the DT binding but does thing slightly differently then.
> > In this case (supporting two incompatible bindings for DT and ACPI), I think
> > the only sensible driver implementation would be to know what we are asking
> > for and use different devm_gpiod_get_index statements based on the firmware
> > interface.
>
> Yes something like that is probably needed.
>
> Alternatively (I didn't try if this works) we could do it so that
> when we see:
>
> gpio = devm_gpiod_get_index(&pdev->dev, "shutdown", 1);
>
> we check first for the property ("shutdown-gpios"), and check if it has
> more than one entry in the value, like:
>
> Package () { "reset-gpios", Package () { ^GPIO, 0, 0, 0, ^GPIO, 1, 0, 0}}
>
> and in that case return the second entry. If we find this instead:
>
> Package () { "reset-gpios", Package () { ^GPIO, 0, 0, 0 }}
>
> we just ignore the index.
>
> Last if there is no _DSD the index refers directly to the GPIO resource
> in _CRS.
>
> This would support both _DSD and non-_DSD at the same time but it makes
> the implementation more complex.
I think the main problem with that approach is that it makes the common
code more error-prone in case of unintentionally broken device descriptions,
because it less often returns an error.
Arnd
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
