Re: [PATCH] amd76x_pm: C3 powersaving for AMD K7

[Erik Slagter <erik@slagter.name>]

[-- Attachment #1: Type: text/plain, Size: 2082 bytes --]

On Fri, 2006-02-03 at 17:59 +0200, Juhani Rautiainen wrote:

> This comes from AMD-768 revision guide. In product errata there there
> is errata number
> 24 which seems to suggest that you can't enable POS, C2 or C3 states
> in single processor
> environments. This at the end of errata:
> ----- snip ----
> This workaround will not work for awaking from the C2/C3 state, since
> the operating
> system has full control. The (ACPI C2/C3) state support is not
> required for Microsoft-compatible workstation and server platforms.
> 
> It is recommended that BIOS should disable the C2 state by
> clearing bit C2EN (device B function 3 offset 4F'h). It is recommended
> that BIOS should disable the C3 state by clearing bit C3EN (device B
> function 3 offset 4F'h).
> ---- snip ----
> 
> If I understand errata correctly then enabling C2/C3 should be safe in
> SMP environment put
> not in single processor enviroment. Maybe module should check this
> before enabling C3.

This is a very cryptic text! What would they mean actually?

That you cannot use POS/C* states in a single processor system OR in SMP
system that only has one processor fitted? Also this bug seems very odd
to me, what would the number of processors have to do with the
availability of C states?

Why "is the C2/C3 state support not required for Microsoft-compatible
platforms"? Do they mean it is okay (in their vision) to have these
running at full power all of the time? Or do they mean there is some
sort of driver that replaces the functionality?

My BIOS (Tyan S2466) does something completely different. It offers an
option "ACPI power saving" (or something like that) in the configuration
screen. This seems to toggle the above mentioned registers. But at the
same time it always marks C2 and C3 unusable by offering impossible
timings. This looks to me that the intention was to have C2/C3 enabled,
but only with this "special" windows driver that does some "extra"
things.

Grmbl, it's all sick.

Does *anyone* have some sort of contact with AMD or Tyan?

[-- Attachment #2: smime.p7s --]
[-- Type: application/x-pkcs7-signature, Size: 2771 bytes --]