On Tue, 2018-02-13 at 08:47 +0100, Ingo Molnar wrote:
> * David Woodhouse <dwmw@amazon.co.uk> wrote:
> 
> > 
> > +extern enum spectre_v2_mitigation spectre_v2_enabled;
>
> This needs to be exported if the KVM module wants to use it.
> 
> > 
> > +static inline bool spectre_v2_ibrs_all(void)
> > +{
> > +	return spectre_v2_enabled == SPECTRE_V2_IBRS_ALL;
> > +}
> > 
> > +     if (vmx->spec_ctrl && !spectre_v2_ibrs_all())
> > 
> > +	if (!spectre_v2_ibrs_all) {
>
> erm, that's a function, not a flag ...

0-day pointed out the latter, which is already fixed in the git tree
ready for the next resend. Not the former though.

I can export it. It does make me ponder for a second whether I should
have gone with my first instinct just to make it another cpufeature
flag like the others. But no, the excuse for doing that for the others
was that it *needs* to be one for using alternatives. And this flag
*isn't* used for alternatives, so it seems a little (more) wrong.