On Fri, Apr 25, 2014 at 12:24:29PM +0100, Peter Maydell wrote: > On 25 April 2014 11:42, Christoffer Dall <christoffer.dall@xxxxxxxxxx> wrote: > > So we're adding an incompletely modelled A57 to the virt board, and > > adding a "board mounted" GICv2. Doesn't sound like something that could > > ever exist in real life, but I agree that the way we are tying things > > together, semantically letting the board decide how to place the GIC > > makes the most sense. > > Actually I think a lot of A57 hardware right now is > using a standalone GIC. The ARM Server Base System > Architecture spec mandates a v2 GIC for level 0/1 > compliance, GICv3 only comes in at level 2. So it's > not that far from reality (and it leaves us the > obvious space to put in a v3 GIC emulation eventually). > > > Oh well. Let's implement GICv3 support some time. > > You first :-) > I hear rumors of people doing this for KVM, so I started refactoring the QEMU GIC code to look more like the in-kernel emulation in terms of style (patches not ready for posting yet), so we can hopefully share some more code later on. -Christoffer _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
