On Sat, 2012-10-27 at 11:01 +0100, Peter Maydell wrote: > This is more or less how ARM has done it (though our specific encoding > of interrupt numbers is different, obviously). > > If I were designing an interface for this kind of thing from scratch > I'd probably want it to look like "create a specific irqchip and give > me some kind of handle to it" and then have an interface for "assert > interrupt line X on that irqchip". Lacking that, a plausible encoding > scheme on the global interrupt numbers works OK given that you know > there aren't going to be that many irqchips in practice... Agreed. It feels like the whole business with routing is not useful for either of us and just adds confusion... Note: I'll be in KVM forum next week (hopefully) so if necessary (and if you guys are there too) we can have a "live" discussion. Cheers, Ben. _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
