Re: [Qemu-devel] [PATCH 3/6] Added Aarch64 CPU Initialization, Get and Put Registers Support.

[Alexander Spyridakis <a.spyridakis@virtualopensystems.com>]

On 29 June 2013 21:17, Peter Maydell <peter.maydell@linaro.org> wrote:
>
> > +#ifdef TARGET_AARCH64
> > +int kvm_arch_init_vcpu(CPUState *cs)
> > +{
> > +    struct kvm_vcpu_init init;
> > +    int ret;
> > +
> > +    /* Try initializing with Foundation Models */
> > +    init.target = KVM_ARM_TARGET_FOUNDATION_V8;
> > +    memset(init.features, 0, sizeof(init.features));
> > +    ret = kvm_vcpu_ioctl(cs, KVM_ARM_VCPU_INIT, &init);
> > +    if (ret) {
> > +        /* Retry initializing with Fast Models */
> > +        init.target = KVM_ARM_TARGET_AEM_V8;
>
> If we're emulating an A57 we should be asking KVM for an A57
> guest. If we're asking KVM for "a CPU like the one in
> the Foundation model" we need to support that via -cpu $something.

Wouldn't that be an overkill at this early stage? There is no target
to test KVM_ARM_TARGET_CORTEX_A57 so we opted for the other two tested
cases.

> But I'm a bit dubious about that anyway -- you need to provide
> a good justification for why KVM/QEMU should be emulating
> emulators and not hardware.

Mainly because there is no real hardware yet. This patch series was
meant to enable basic working support of KVM in AArch64 on what is
available and open discussion on how to proceed from there. Also,
given that fact that kvmtool uses the same approach it was the most
obvious way at the time to handle it.