RE: [PATCH 1/2] amd_nb: Add PCI ID for family 19h model 78h

[Limonciello, Mario]

[Public]

> On Thu, Apr 27, 2023 at 12:33:36AM -0500, Mario Limonciello wrote:
> > s2idle previously worked on this system, but it regressed in kernel
> > 6.4 due to commit 310e782a99c7 ("platform/x86/amd: pmc: Utilize SMN
> > index 0 for driver probe").
> >
> > The reason for the regression is that before this commit the SMN
> > communication was hardcoded, but after amd_smn_read() is used which
> > relies upon the misc PCI ID used by DF function 3 being included in
> > a table.  The ID was missing for model 78h, so this meant that the
> > amd_smn_read() wouldn't work.
> >
> > Add the missing ID into amd_nb, restoring s2idle on this system.
> 
> Is there a long-term solution for this that will not require adding
> new IDs every time new hardware comes out?
> 
> drivers/platform/x86/amd/pmc.c already matches ACPI IDs; maybe there's
> some way for the platform to provide the information you need via
> ACPI or something?
> 
> Bjorn

I had the same question when I came up with this and found out the 
ACPI tables don't encode the information needed right now.

But, yes there are discussions ongoing to make this more scalable in
a future generation.