Re: ACPI call for testing

[Jukka Ruohonen <jruohonen%iki.fi@localhost>]

On Sun, Sep 05, 2010 at 09:41:28AM +0100, David Laight wrote:
> On Wed, Aug 25, 2010 at 05:15:44PM +0200, Christoph Egger wrote:
> > 
> > The correct workaround is to
> > a) implement the OSVW feature
> > b) use OSVW to detect Errata 400. When present then switch away
> >    from the LAPIC timer before entering HLT.
> 
> Doesn't that just mean the LAPIC timer is completely unusable?

Yes.

> Since the HLT (as 'wait for interrupt') happens every time the idle
> loop is entered and it isn't sane to change the clock source that often.

Few general ideas about different approaches: (a) do not use the LAPIC timer
at all but instead resort to per-CPU timers (HPET?), (b) proxy the stalled
LAPIC via HPET or some other timer (as suggested above), or (c) ensure that
one processor never enters any idle-state and use that CPU to wake other
idling CPUs using IPI.

- Jukka.