Re: x86-fix-cmpxchg.patch added to -mm tree

[Denis Vlasenko <vda@ilport.com.ua>]

On Monday 12 September 2005 09:41, Jan Beulich wrote:
> >>> Dave Jones <davej@redhat.com> 10.09.05 02:30:38 >>>
> >On Fri, Sep 09, 2005 at 05:06:56PM -0700, Andrew Morton wrote:
> >
> > > - cmpxchg8b gets disabled when the minimum specified hardware
> architectur
> > >   doesn't support it (like was already happening for the byte,
> word, and
> > >   long ones).
> > >
> > > +config X86_CMPXCHG64
> > > +	bool
> > > +	depends on !M386 && !M486 && !MCYRIXIII && !MGEODEGX1
> > > +	default y
> >
> >This is wrong.  All the Winchip/CyrixIII CPUs do indeed have cx8.
> >Though cpuid will report that it's missing until We explicitly enable
> >it in init_c3().  Whilst it's "disabled" however, the instruction does
> >run perfectly fine.   This was done because Win NT wouldn't boot if it
> >found a non Intel CPU which had cx8 iirc.

Can this be added to comments in the right place?

> >I'm not familiar with the Geode enough to answer definitively.
> >Alan seems to be the Geode-guru, and may still have datasheets for
> that.
> 
> Resubmitting adjusted patch (taking into account also Alan's later
> response).
--
vda