Re: [GIT PULL][RESEND #2] KVM Updates for 2.6.23-rc1

["H. Peter Anvin" <hpa@zytor.com>]

Linus Torvalds wrote:
> 
> Actually, I think the *real* solution would be:
> 
>  - add a X86_HAS_CMPXCHG8B config option, and set it for the appropriate 
>    CPU selection (P6 and up, or whatever the rule is)
> 
>  - make KVM depend on it
> 
>  - make KVM and HIGHMEM64 _select_ another config option, namely the 
>    NEEDS_CMPXCHG8B
> 
> and then we make the cpufeatures code check the CMPXCHG bit only if the 
> NEEDS_CMPXCHG8B thing is set. That gives us the best of all worlds.
> 
> Because there is no point in checking whether the CPU supports it if the 
> kernel doesn't _need_ it. Especially since we know that some CPU's lie 
> about it due to old NT bugs.
> 

Yes.  This is probably also true for other CPU feature options, at least
in the abstract (i.e. it may not apply to any of the current options).

	-hpa