Re: [PATCH v8 2/2] kvm: x86: hyperv: guest->host event signaling via eventfd

["Radim Krčmář" <rkrcmar@redhat.com>]

2018-02-01 15:45+0300, Roman Kagan:
> On Wed, Jan 31, 2018 at 07:18:12PM +0100, Radim Krčmář wrote:
> > 2018-01-31 16:56+0300, Roman Kagan:
> > I have completely missed this series in the past, sorry.
> 
> No wonder, everybody was too busy with all this Meltdown+Spectre
> stuff.
> 
> > > +	/*
> > > +	 * Per spec, bits 32-47 contain the extra "flag number".  However, we
> > > +	 * have no use for it, and in all known usecases it is zero, so just
> > > +	 * require it here.
> > > +	 */
> > 
> > The spec says:
> > 
> >   FlagNumber specifies the relative index of the event flag that the
> >   caller wants to set within the target SIEF area. This number is
> >   relative to the base flag number associated with the port.
> 
> This smells like an overdesign to me.  Anyway we've never seen non-zero
> FlagNumbers.  The guest code in Linux doesn't even define a way to pass
> such a FlagNumber.
> 
> > Don't we want to pass the flag into userspace?
> 
> To be future-proof in case there turns out to be a user for non-zero
> FlagNumber?  Makes sense, will respin with that.

Exactly, thanks!