On Tue, Sep 19, 2017 at 04:52:10PM +0200, Cédric Le Goater wrote:
> On 09/19/2017 04:45 AM, David Gibson wrote:
> > On Mon, Sep 11, 2017 at 07:12:19PM +0200, Cédric Le Goater wrote:
> >> The number of IPIs is deduced from the max number of CPUs the guest
> >> supports and the IRQ numbers for the IPIs are allocated from the top
> >> of the IRQ number space to reduce conflict with other IRQ numbers
> >> allocated by the devices.
> >>
> >> Signed-off-by: Cédric Le Goater <clg@kaod.org>
> > 
> > This is more ick associated with implementing XIVE in terms of XICS.
> > We shouldn't need to "allocate" IRQs for the IPIs - they should just
> > be a fixed set.  
> 
> They are allocated at the right beginning so we can consider them
> fixed I suppose. 
> 
> > And we certainly shouldn't need to set the XICS irq type for XIVE irqs.
> 
> This is because, in this patchset, XIVE and XICS use the same IRQ 
> allocator which happens to be the ICSIRQState array of XICS. yes, 
> this is ugly but we are identifying the different constraints. 

Yeah, as I said in the other mail, I think trying to support both
immediately is making a mess of the XIVE design.  Let's get it working
as a machine option first, then worry about CAS and migration.

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson