On Tue, Sep 21, 2021, Vitaly Kuznetsov wrote:
> Sean Christopherson <seanjc@xxxxxxxxxx> writes:
>
> > Move the few bits of relevant fx_init() code into kvm_arch_vcpu_create(),
> > dropping the superfluous check on vcpu->arch.guest_fpu that was blindly
> > and wrongly added by commit ed02b213098a ("KVM: SVM: Guest FPU state
> > save/restore not needed for SEV-ES guest").
>
> I have more questions to the above mentioned commit: why is it OK to
> 'return 0' from kvm_vcpu_ioctl_x86_set_xsave() without writing anything
> to 'guest_xsave'? Same goes to kvm_arch_vcpu_ioctl_get_fpu(). Whould't
> it be better to throw an error as we can't actually get this information
> for encrypted guests? It's probably too late to change this now I
> suppose ...
Yep, I would also have preferred that KVM force userspace to avoid ioctls() that
cannot work, but that ship has sailed.
