Comment 618 for bug 1690085

(In reply to Klaus Mueller from comment #530)
> Hmm, if rev. 1 doesn't support MWAIT - why can it be a problem anyway
> at the same time which must be fixed by disabling the usage of MWAIT?
> I seem to miss something?

That's a good question but, frankly, I don't have a very exact answer to
it right now.

In order to understand what's *really* going on in the
cstate detection code one would need to instrument
at least acpi_processor_get_power_info_cst() and
acpi_processor_ffh_cstate_probe() to figure out what exactly does the
kernel parse from those CST objects and what it uses to try to enter
idle.

And do all that instrumentation on an affected system.

My current suspicion is is that it tries to enter idle with
misconfigured states and under certain conditions, it misses the wakeup,
leading to the stall.

This is all a conjecture anyway.

Now my patch simply falls back to the good old idle entry on AMD where
we simply do HLT and we won't even attempt to enter idle the ACPI way.

Makes sense?