Re: [PATCH v5 1/1] x86: kvm: svm: set up ERAPS support for guests

[Amit Shah <amit@kernel.org>]

On (Wed) 20 Aug 2025 [09:18:29], Sean Christopherson wrote:
> On Wed, May 28, 2025, Amit Shah wrote:
> > On Mon, 2025-05-19 at 16:22 -0500, Tom Lendacky wrote:
> > > > +static inline void vmcb_set_flush_guest_rap(struct vmcb *vmcb)
> > > > +{
> > > > +	vmcb->control.erap_ctl |= ERAP_CONTROL_FLUSH_RAP;
> > > > +}
> > > > +
> > > > +static inline void vmcb_clr_flush_guest_rap(struct vmcb *vmcb)
> > > > +{
> > > > +	vmcb->control.erap_ctl &= ~ERAP_CONTROL_FLUSH_RAP;
> > > > +}
> > > > +
> > > > +static inline void vmcb_enable_extended_rap(struct vmcb *vmcb)
> > > 
> > > s/extended/larger/ to match the bit name ?
> > 
> > I also prefer it with the "larger" name, but that is a confusing bit
> > name -- so after the last round of review, I renamed the accessor
> > functions to be "better", while leaving the bit defines match what the
> > CPU has.
> > 
> > I don't mind switching this back - anyone else have any other opinions?
> 
> They both suck equally?  :-)
> 
> My dislike of "larger" is that it's a relative and intermediate term.  What is
> the "smaller" size?  Is there an "even larger" or "largest size"?
> 
> "extended" doesn't help in any way, because that simply "solves" the problem of
> size ambiguity by saying absolutely nothing about the size.

I agree with that; but it's just how the bit is named in the APM, so...

> I also dislike "allow", because in virtualization context, "allow" usually refers
> to what the _guest_ can do, but in this case "allow" refers to what the CPU can
> do.

(same as above)

> If we want to diverge from the APM, my vote would be for something like
> 
>   ERAP_CONTROL_FULL_SIZE_RAP

Oh I def didn't want to diverge from the APM (for the name of the bit).  I
only wnat to check what's palatable for the accessors -- but a quick read of
the followup reply shows you're asking to drop them and just use the checks
directly, that's fine - no need for these accessors and this renaming.

	  	 Amit