Re: [PATCH 2/2] platform/x86/intel/pmc: Add PSON residency counter

[Andy Shevchenko]

On Sat, Jul 24, 2021 at 10:49 PM David E. Box
<david.e.box@xxxxxxxxxxxxxxx> wrote:
> On Sat, 2021-07-24 at 11:29 +0300, Andy Shevchenko wrote:
> > On Friday, July 23, 2021, Michael Bottini
> > <michael.a.bottini@xxxxxxxxxxxxxxx> wrote:
> > > Tiger Lake devices have the capability to track the duration
> > > of time that their Power Supply Units (PSUs) are turned off during
> > > S0ix.
> > > This patch adds a debugfs file `pson_residency_usec` to provide
> > > access to this counter.
> > >
> > > In order to determine whether the device is capable of PSON,
> > > use acpi_init_properties() to reevaluate _DSD.
> > >
> > >
> >
> > It’s direct abuse of ACPI specification as I read it:
> >
> > “_DSD must return the same data each time it is evaluated. Firmware
> > should not expect it to be evaluated every time (in case it is
> > implemented as a method).”
> >
> >
> > NAK to the series.
>
> Okay, we'll check with the BIOS folks. They are setting this property
> from _STA. They may not expect OSPM to reevaluate _DSD.

Does it matter?

> But they may
> have expected that OSPM doesn't attempt to read _DSD until after _STA
> is executed.

They may study a bit of ACPI specification perhaps?
Or if there is indeed an issue with the specification language (so
native-speaking people misinterpret above) then ECR to ASWG should be
submitted?

-- 
With Best Regards,
Andy Shevchenko