On 8 March 2017 at 18:21, Christoffer Dall <christoffer.dall@xxxxxxxxxx> wrote: > On Wed, Mar 08, 2017 at 03:27:28PM +0100, Peter Maydell wrote: >> From a userspace API point of view, I was expecting this to be >> "just don't do that then" territory, ie it's userspace's job >> to set the MPIDRs to something that makes sense. Does KVM >> actually care internally what the MPIDR presented to the guest is? >> (In theory I don't see why you shouldn't be able to present >> the guest with a bit of bogus hardware that claims the same >> MPIDR for all cores.) >> > > I think it's important that we maintain a strict mapping between an > MPIDR and a VCPU at any given time, because otherwise, we'll have > problems knowing which redestributor belongs to which VCPU and making > other qualified decisions in the KVM/ARM implementation. Mmm, maybe. In any case I certainly don't have an objection if you want to define the interface as "must set before VCPU start and not after" and fail VCPU start if it's using a duplicate ID. thanks -- PMM _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
